Predicting abnormal returns from news using text classification

We show how text from news articles can be used to predict intraday price movements of financial assets using support vector machines. Multiple kernel learning is used to combine equity returns with text as predictive features to increase classification performance and we develop an analytic center cutting plane method to solve the kernel learning problem efficiently. We observe that while the direction of returns is not predictable using either text or returns, their size is, with text features producing significantly better performance than historical returns alone.     

Estimating GARCH models using support vector machines*

Abstract

Support vector machines (SVMs) are a new nonparametric tool for regression estimation. We will use this tool to estimate the parameters of a GARCH model for predicting the conditional volatility of stock market returns. GARCH models are usually estimated using maximum likelihood (ML) procedures, assuming that the data are normally distributed. In this paper, we will show that GARCH models can be estimated using SVMs and that such estimates have a higher predicting ability than those obtained via common ML methods.     

Time-varying neural network for stock return prediction

We consider the problem of neural network training in a time-varying context. Machine learning algorithms have excelled in problems that do not change over time. However, problems encountered in financial markets are often time varying. We propose the online early stopping algorithm and show that a neural network trained using this algorithm can track a function changing with unknown dynamics. We compare the proposed algorithm to current approaches on predicting monthly US stock returns and show its superiority. We also show that prominent factors (such as the size and momentum effects) and industry indicators exhibit time-varying predictive power on stock returns. We find that during market distress, industry indicators experience an increase in importance at the expense of firm level features. This indicates that industries play a role in explaining stock returns during periods of heightened risk.     

Do stock bulletin board systems (BBS) contain useful information? A viewpoint of interaction between BBS quality and predicting ability

This study explores whether information on internet stock bulletin board systems (BBS) is valuable for stock return prediction, taking advantage of data derived from the biggest stock BBS in China. Using a text classification algorithm, we find the online messages significantly predict stock return with negligible R‐squared. However, we find that accuracy of individual BBS posts is below 50 percent and there is no distinction at prediction accuracy between high‐ and low‐quality stock BBS. Due to the autocorrelation of stock returns, we argue that BBS predicts stock returns because of its reflection on the simultaneous stock return rather than revelation on valuable information.     

Dynamic Interactions among the Stock Market,Federal Funds Rate,Inflation, and Economic Activity

This paper examines the dynamic interactions among the equity market, economic activity, inflation, and monetary policy under three monetary policy regimes using bivariate and multivariate vector autoregressive cointegrating specifications. The bivariate results for the real stock returns‐inflation pair weakly support a negative correlation in the 1970s and 1980s. While the bivariate findings suggest a weak, negative relationship between real returns and the federal funds in the 1970s and 1980s, the multivariate findings strongly support short‐term linkages in the 1970s. There appears to be no consistent dynamic relationship between monetary policy and stock prices in that the relationship differs across monetary regimes.     

MEASURING EFFECTS ON STOCK RETURNS OF SENTIMENT INDEXES CREATED FROM STOCK MESSAGE BOARDS

Various techniques and sources of information exist to aid investors in predicting future stock returns. However, no effective proxy for retail investors, such as stock message board users, has been established. This study provides guidelines for creating an effective proxy. The heart of such proxies is sentiment indexes, and in the past the indexes have had low predictive power. Introducing four methodological improvements for applying text classifiers and two probability measurements, we contrast eight widely applied text classifiers to stock message board data. Based on the classifier results and incorporating our new methods, the new sentiment index proves to be a significant “same‐day positive but next‐day negative” directional indicator.     

The change in stock-selection risk and stock market returns

Following Jiang et al. (2021), who propose a stock-selection opportunity (SSO) measurement by the absolute average positive alpha of individual stocks to reflect stock-selection timing, we construct a stock-selection risk (SSR) measure from the perspective of negative alphas of individual stocks. Then, we investigate the predictive abilities of SSO, SSR, the change of SSO (CSSO), and the change of SSR (CSSR) on stock market returns. By using data from 2003 to 2020 in China, we find that only CSSR significantly predicts future one-month market returns. Moreover, considering other popular predictors, our in-sample and out-of-sample results and a series of robustness checks support the proposal that CSSR provides unique information for predicting market returns that reduces forecast errors and increases economic value for investors. Furthermore, our trading activity evidence shows that CSSR predicts stock market returns due to investors' underreaction to the information of CSSR.     

Performance-weighted ensembles of random forests for predicting price impact

For any large player in financial markets, the impact of their trading activity represents a substantial proportion of transaction costs. This paper proposes a novel machine learning algorithm for predicting the price impact of order book events. Specifically, we introduce a prediction system based on ensembles of random forests (RFs). The system is trained and tested on depth-of-book data from the BATS and Chi-X exchanges and performance is benchmarked using ensembles of other popular regression algorithms including: linear regression, neural networks and support vector regression. The results show that recency-weighted ensembles of RFs produce over 15% greater prediction accuracy on out-of-sample data, for 5 out of 6 timeframes studied, compared with all benchmarks. Feature importance ranking is used to explore the significance of various market features on the price impact, finding them to be highly variable through time. Finally, a novel procedure for extracting the directional effects of features is proposed and used to explore the features most dominant in the price formation process.     

Statistical arbitrage with optimal causal paths on high-frequency data of the S&P 500

This paper develops the optimal causal path algorithm and applies it within a fully-fledged statistical arbitrage framework to minute-by-minute data of the S&P 500 constituents from 1998 to 2015. Specifically, the algorithm efficiently determines the optimal non-linear mapping and the corresponding lead–lag structure between two time series. Afterwards, this study explores the use of optimal causal paths as a means for identifying promising stock pairs and for generating buy and sell signals. For this purpose, the established trading strategy exploits information about the leading stock to predict future returns of the following stock. The value-add of the proposed framework is assessed by benchmarking it with variants relying on classic similarity measures and a buy-and-hold investment in the S&P 500 index. In the empirical back-testing study, the trading algorithm generates statistically and economically significant returns of 54.98% p.a. and an annualized Sharpe ratio of 3.57 after transaction costs. Returns are well superior to the benchmark approaches and do not load on any common sources of systematic risk. The strategy outperforms in the context of cryptocurrencies even in recent times due to the fact that stock returns contain substantial information about the future bitcoin returns.     

What Moves the Stock and Bond Markets? A Variance Decomposition for Long-Term Asset Returns

This paper uses a vector autoregressive model to decompose excess stock and 10-year bond returns into changes in expectations of future stock dividends, inflation, short-term real interest rates, and excess stock and bond returns. In monthly postwar U.S. data, stock and bond returns are driven largely by news about future excess stock returns and inflation, respectively. Real interest rates have little impact on returns, although they do affect the short-term nominal interest rate and the slope of the term structure. These findings help to explain the low correlation between excess stock and bond returns.     

Revisiting the empirical linkages between stock returns and trading volume

This paper investigates whether the empirical linkages between stock returns and trading volume differ over the fluctuations of stock markets, i.e., whether the return–volume relation is asymmetric in bull and bear stock markets. Using monthly data for the S&P 500 price index and trading volume from 1973M2 to 2008M10, strong evidence of asymmetry in contemporaneous correlation is found. As for a dynamic (causal) relation, it is found that the stock return is capable of predicting trading volume in both bear and bull markets. However, the evidence for trade volume predicting returns is weaker.     

On the real estate market efficiency

This reseach reexamines the efficiency hypothesis of the real estate market using monthly data and the vector autoregressive (VAR) modelling technique. The tests focus on the causal linkage between real estate returns and a number of relevant financial and economic variables. An eight-by-eight VAR model is estimated using the FPE and the specific gravity criteria, in conjunction with an extensive series of specification tests. The empirical results distilled from system estimations suggest that the real estate market is efficient with respect to available information on the industrial production, the risk premia, the term structure of interest rates, and the monetary base. Movements in these variables are quickly and fully utilized by market agents, perhaps owing to the intensity with which their relationship with stock returns has been discussed in the literature and the popular media. However, the results also suggest the presence of a significant lagged relationship between real estate returns and fiscal policy moves, even when the paths through other potential determinants of these returns are taken into account. Of course, our finding that the fiscal policy measure is useful in predicting stock returns does not necessarily imply that the real estate market is inefficient. At a minimum, inefficiency is revealed only if a careful analysis of the budgetary process can help design a profitable (exploitable) trading strategy.     

Nonlinear tail dependence in cryptocurrency-stock market returns: The role of Bitcoin futures

We investigate the median and tail dependence between cryptocurrency and stock market returns of BRICS and Developed countries using a newly developed nonparametric cumulative measure of dependence over the period January 4, 2016 – December 31, 2019 as well as before and after the introduction of Bitcoin futures on December 17, 2017. The new measure is model-free and permits measuring tail risk. The results highlight the leading role of S&P500, Nasdaq and DAX 30 in predicting BRICS and developed countries’ stock market returns. Among BRICS countries, BVSP shows a starring role in predicting stock market returns. BSE 30 is the most predictor of cryptocurrencies, which have a little predictability on stock market returns. Ethereum has the leading role in predicting cryptocurrencies and stock market returns followed by Bitcoin. Tail dependence shows substantial role of S&P500, Nasdaq and BVSP in predicting stock market returns. Subsample analysis show the role of Bitcoin futures in reshaping the mean and tail dependence between cryptocurrency and stock market returns. Our results have important policy implications for portfolio managers, hedge funds and investors.     

The Pricing of Relative Performance Based Incentives for Executive Compensation

Japanese firms report both parent-only and consolidated financial statements. Because of the unique business environment in Japan, there is a widely held view that parent-only data provides a better means for assessing the value of the entire firm. We find that both parent-only and subsidiary earnings are important in predicting future consolidated earnings. However, while stock prices accurately reflect the persistence of parent-only earnings, the Japanese stock market appears to underestimate the persistence of subsidiary earnings, causing a significant positive relation between changes in subsidiary earnings in year t and stock returns in year t +1. This relation between subsidiary earnings and future stock returns does not persist beyond year t 7plus;1. Taking a long (short) position in firms with large, positive (negative) changes in subsidiary earnings results in an average annual abnormal return of 7.06% with positive returns in 12 of the 13 years in the test period.     

Linkages among European and world stock markets

Causal relations and dynamic interactions among equity returns in ten countries for the period 1983–1994 are analysed. An innovation accounting approach based on a multivariate vector autoregressive (VAR) model is used to estimate the proportion of each market return's forecast error attributable to innovations in foreign market returns. Three major results appear. The variance decompositions indicate a strong degree of economic interaction among stock markets. The US stock market has a considerable influence on stock market performance in almost every country, while there is no substantial inter-continental influence from the European stock markets on the world's two largest equity markets in New York and Tokyo. Finally, the pattern of the impulse-response functions illustrates a rapid international transmission of stock market events, supporting the hypothesis of international stock market efficiency.     

Causal and Dynamic Relationships among Stock Returns,Return Volatility and Trading Volume: Evidence from Emerging markets in South-East Asia

This paper examines the causal and dynamic relationships among stock returns, return volatility and trading volume for five emerging markets in South-East Asia—Indonesia, Malaysia, Philippines, Singapore and Thailand. We find strong evidence of asymmetry in the relationship between the stock returns and trading volume; returns are important in predicting their future dynamics as well as those of the trading volume, but trading volume has a very limited impact on the future dynamics of stock returns. However, the trading volume of some markets seems to contain information that is useful in predicting future dynamics of return volatility.     

Are Japanese margin buyers informed?

Using Japanese long sample (1977–2010) market data, we examine whether margin buying is informed trades about future stock returns and whether they are related to undervaluation of the market. We find that margin buying increases when temporary returns are higher contemporaneously. We do not find that Japanese margin buying is well-informed in predicting future permanent changes in stock returns. Further, we find that margin buying is not related to the undervaluation of stock market prices.     

The dynamic interaction between volatility and returns in the US stock market using leveraged bootstrap simulations

One of the most important stylized facts in finance is that stock index returns are inversely related to volatility. The theoretical rationale behind the proposition is still controversial. The causal relationship between returns and volatility is investigated in the US stock market over the period 2004-2009 using daily data. We apply a bootstrap test with leveraged adjustments that is robust to non-normality and ARCH. We find that the volatility causes returns negatively and returns cause volatility positively. The policy implications of our findings are discussed in the main text.     

Revisiting calendar anomalies in Asian stock markets using a stochastic dominance approach

Extensive evidence on the prevalence of calendar effects suggests that there exist abnormal returns. Some recent studies, however, have concluded that calendar effects have largely disappeared. In spite of the non-normal nature of stock returns, most previous studies have employed the mean-variance criterion or CAPM statistics. These methods rely on the normality assumption and depend only on the first two moments to test for calendar effects. A limitation of these approaches is that they miss important information contained in the data such as higher moments. In this paper we use a stochastic dominance (SD) test to test for the existence of day-of-the-week and January effects. We use daily data for 1988–2002 for several Asian markets. Our empirical results support the existence of weekday and monthly seasonality effects in some Asian markets, but suggest that first-order SD for the January effect has largely disappeared.     

Stock returns,aggregate earnings surprises,and behavioral finance

We study the stock market's reaction to aggregate earnings news. Prior research shows that, for individual firms, stock prices react positively to earnings news but require several quarters to fully reflect the information in earnings. We find a substantially different pattern in aggregate data. First, returns are unrelated to past earnings, suggesting that prices neither underreact nor overreact to aggregate earnings news. Second, aggregate returns correlate negatively with concurrent earnings; over the last 30 years, for example, stock prices increased 5.7% in quarters with negative earnings growth and only 2.1% otherwise. This finding suggests that earnings and discount rates move together over time and provides new evidence that discount-rate shocks explain a significant fraction of aggregate stock returns.