Sentiment Measures Research Articles

Benefiting from the frightened herd: Dynamic asset allocation amid panic sentiment

This paper introduces an innovative dynamic asset allocation approach to hedge against panic sentiment and demonstrates that adopting such an anti-panic strategy can be beneficial. We first develop a novel measure of panic sentiment and analyze its spillover effects across major stock markets. The strategy’s diversification advantages in mitigating panic sentiment are then evaluated. Backtesting results reveal substantial long-term excess returns over benchmarks, with risk-adjusted metrics and statistical tests confirming the robustness of these results.

Sentiment analysis of economic text: A lexicon‐based approach

AbstractWe propose an economic lexicon (EL) specifically designed for textual applications in economics. We construct the dictionary with two important characteristics: (1) to have a wide coverage of terms used in documents discussing economic concepts, and (2) to provide a human‐annotated sentiment score in the range [−1, 1]. We illustrate the use of the EL in the context of a simple sentiment measure and consider several applications in economics. The comparison to other lexicons shows that the EL is superior due to its wider coverage of domain relevant terms and its more accurate categorization of the word sentiment.

Instability of discount-rate and cash-flow channels of stock returns: a nonparametric approach

ABSTRACT According to the log-linear return approximation, the ability of a predictor to predict future stock returns may arise from its ability to predict either the cash flows or the discount rates, or both. This paper introduces a novel nonparametric approach to investigating the time variability in the predictability channels of the equity premium within a present-value framework. Using the existing return predictors, including the measures of economic uncertainty and sentiment, we document strong time variation in both discount-rate and cash-flow channels. In comparison, the predictability of the equity premium through the cash-flow channel is more pronounced than through the discount-rate channel.

Realized Stock Market Volatility of the United States: The Role of Employee Sentiment

Using data for the United States (US) stock market covering the sample period from 2008:06 to 2020:12, we study the incremental predictive value of employee sentiment for the realized volatility of stock returns. In doing so, we control four different measures of investor sentiment and various macroeconomic and financial factors and uncertainties. We report results for several combinations of forecast horizons and estimation windows and find that employee sentiment contributes to forecast accuracy for several combinations.

Auditor mood and audit quality: evidence from music sentiment measure

PurposeExtant studies propose music sentiment as a novel measure of individuals’ sentiment. These studies argue that individuals’ choice of music reflects their emotional condition in real time and influences their cognitive ability, making it a powerful tool for assessing their mood. This study aims to use music sentiment as a proxy for auditors’ mood and explore its impact on audit quality.Design/methodology/approachA sample of the US firms from 2017 to 2020 is used in the study. The authors apply the ordinary least squares regressions and the logit regressions to the audit quality models. The authors use absolute discretionary accruals and the propensity to meet or beat earnings forecasts as proxies for audit quality and calculate a stream-weighted average sentiment measure for Spotify’s Top-200 songs of each day during the audit period of a client firm to capture the sentiment of auditors.FindingsThe authors find that music sentiment is positively associated with audit quality. The result is consistent with the mood maintenance hypothesis, which suggests that a positive mood can induce auditors to be more careful in risky situations. Furthermore, the result is robust to various sensitivity analyses.Originality/valueThe study contributes to the scarce literature that focuses on auditors’ emotional state and highlights the importance of monitoring auditor mindset during the audit period.

Carbon credit sentiments and green energy stocks

ABSTRACT This paper provides a new perspective on the role of climate attention in the media for returns of green energy stocks. Making use of a data set that includes various measures of sentiments and media attention in newspapers and social media related to 1) carbon credit and 2) stock markets, we disentangle the drivers of green stock returns. Our results for various green stock indices suggest that green stock returns are not only driven by stock-related news but also driven by sentiment related to carbon credit news. We also show that the underlying news source (social media vs. newspapers) matters. News related to uncertainty around carbon credit in conventional newspapers is accompanied by decreasing returns. In social media, future-oriented pessimistic emotions relate negatively and recent positive sentiment relates positively to green stock returns.

Bitcoin Sentiment Analysis using Machine Learning Algorithms

The rapid rise of cryptocurrencies, particularly Bitcoin, has intensified research interest in understanding the factors influencing their price movements. Among these factors, social media sentiment has emerged as a crucial predictor, reflecting collective investor mood and market expectations. This paper provides a comprehensive survey of various sentiment analysis models applied in cryptocurrency markets, with a specific focus on the relationship between social media sentiment and Bitcoin price fluctuations. The study identifies the Aigents model as the most effective, showing significant improvements in predictive accuracy following fine-tuning. Findings reveal a predictive association between sentiment measures and price changes, typically with a latency of one to two days. The paper offers insights into the capabilities and limitations of existing Natural Language Processing (NLP) models in the context of cryptocurrency sentiment analysis, presenting practical implications for investors and analysts in navigating the volatile cryptocurrency markets. Key Words: cryptocurrency, Bitcoin, social media sentiment, natural language processing (NLP), sentiment analysis, price prediction, machine learning models, Aigents model, financial forecasting, Twitter, Reddit, artificial neural networks (ANN), Long Short-Term Memory (LSTM), Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Random Forest, Naive Bayes, Extreme Gradient Boosting (XGBoost), predictive analytics, market behavior, interpretable AI.

Analysis of the Impact of Investor Sentiment on Stock Price Crash Risk

The stability of the stock market is very important to the country. Some researchers have found that sentiment has an impact on stock price crash risk. However, research mainly focuses on media sentiment and company sentiment, and research on investor sentiment is not mature enough. This article collates the literature to explore the relationship between investor sentiment and stock price crash risk. This article provides a systematic classification of indicators that measure stock price crash risk. Then, when introducing the methods of measuring investor sentiment, the indicators for measuring investor sentiment are classified, and the scope of application of the three mainstream methods is summarized. This article introduces three models commonly used in the academic community for this problem and conducts a comparative analysis of the three mainstream models. Research has found a positive relationship between investor sentiment and stock price crash risk. As crisis sentiment rises, a large number of investors will begin to implement stop-loss strategies and even sell at extremely low prices to minimize losses. As a result, the likelihood of a stock market price crash increases significantly. In addition, the selection of data sets affects the measurement of investor sentiment to a great extent, and the development and utilization of alternative data is a future research direction. Beyond this, there is reverse causation between investor sentiment and stock price crash risk. Therefore, it is necessary to consider reducing or avoiding the impact of endogeneity problems when selecting a model.

Do optimistic portfolios outperform pessimistic portfolios: Evidence from textual sentiment

We examine whether textual sentiment (from news and social media) explains the cross-section of stock returns. Sentiment scores are used to sort stocks into tercile portfolios daily. Various asset pricing models, including CAPM and Fama-French models, are used to assess the sentiment's impact on returns after accounting for traditional risk factors. Results show that portfolios with higher (optimistic) sentiment consistently yield better average returns across the two sentiment measures and the different asset pricing models.

Predicting stock market crashes in MENA regions: study based on the irrationality of investor behavior and the NARX model

PurposeThis study aims to predict stock market crises in the Middle East North Africa (MENA) regions by leveraging the nonlinear autoregressive neural network with exogenous inputs (NARX) model with two measures of investor sentiment: the ARMS indicator and Google Trends' search volume of positive and negative words.Design/methodology/approachEmploying a novel approach, this study utilizes the NARX model with ten neurons in the hidden layer and the Levenberg–Marquardt training algorithm. It evaluates model performance through learning, validation and test errors, as well as correlation analysis between predicted and actual crises.FindingsThe NARX model, incorporating investor sentiment, has proven to be a reliable tool for forecasting crises, helping market participants understand data complexity and avoid crisis consequences. The divergence in how investors interpret market news, with some focusing solely on negative developments and others valuing positive outcomes, highlights the predictive nature of the optimistic and pessimistic sentiments captured by the model.Research limitations/implicationsThis study advocates for integrating behavioral approaches into stock market crisis prediction, highlighting the significance of investor sentiment and deep learning. It advances crisis mechanism understanding and opens avenues in behavioral finance. Integration of these findings into finance and economics education could enhance students' risk understanding and mitigation strategies.Practical implicationsThe adoption of NARX models, incorporating investor sentiment, empowers market participants to proactively manage crises, adjust strategies, enhance asset protection and make informed decisions. These models enable them to minimize losses, maximize returns and diversify portfolios effectively in response to market fluctuations. These insights also guide policymakers such as governments, regulatory institutions and financial organizations in formulating crisis prevention and mitigation policies, bolstering economic and financial stability.Social implicationsThis research reduces economic uncertainty, safeguards individuals' savings and investments and promotes a stable financial climate.Originality/valueThis study is one of the first attempts to demonstrate the detection and prediction of stock market crises, specifically in the MENA stock market, using the NARX model. It offers a robust forecasting model using machine learning and investor sentiment, providing decision-making support for investment strategies and policy development aimed at enhancing financial and economic stability.

The role of uncertainty and sentiment for intraday volatility connectedness between oil and financial markets

We quantify intraday volatility connectedness between oil and key financial assets and assess how it is related to uncertainty and sentiment measures. For that purpose, we integrate the well-known spillover methodology with a TVP VAR model estimated on a unique, vast dataset of roughly 300 thousand 5 min quotations for most heavily traded financial assets: crude oil, the US dollar, S&P 500 index, gold and US treasury bonds. This distinguishes our investigation from previous studies, which usually employ relatively short samples of daily or weekly data and focus on connectedness between two asset classes. We contribute to the literature across three margins. First, we document that market connectedness at intraday frequency presents a different picture on markets co-movement compared to the estimates obtained using daily data. Second, we show that at 5 min frequency volatility is mostly transmitted from the stock market and absorbed by the bond and dollar markets, with oil and gold markets being occasionally important for volatility transmission. Third, we present evidence that daily averages of intraday connectedness measures respond to changes in sentiment and market-specific uncertainty. Interestingly, our results contrast with earlier findings, as they show that connectedness among markets decreases in periods of high volatility owing to market-specific factors. Our study points to the importance of using high-frequency data in order to better understand financial and commodity markets dynamics.

Fair value estimates for illiquid cryptocurrency

To address the need for reporting and disclosure of cryptocurrency holdings in compliance with the FASB guidance for the use of fair value measurements for cryptocurrency (FASB, 2023), this paper develops a modeling process for reporting entities to measure the market value of cryptocurrencies with limited or no observable transactions. In this valuation model, we consider the last observable market information with time decay, its comparable assets market index, and dynamic real-time market participants’ sentiment and attention. Notably, the application of exogenous variables allows us to maximize the observable inputs in measuring fair value, such as asset classification based on economic traits and market participants’ attention and sentiment measurement with online media textual analytics. We propose a valuation framework and construct a prediction model that can achieve a prediction accuracy of 87 % on target asset resurging prices.

Examining Implications of Search Query Construction on Sentiment Analysis: The Case of Brandwatch

Consistent with other fields, advertising research is currently facing a replication crisis, evoking methodological concerns regarding the reproducibility of research findings. To address this crisis, we systematically reviewed seven computational advertising studies that collected and analyzed the sentiment of tweets. Upon identifying five studies that reported replicable data collection processes, we utilized the original search queries to recollect tweets using a third-party platform, Brandwatch. We also replicated data collection using iterative complex Boolean (ICB) operators, which sought to enhance the resulting dataset precision. Then, we juxtaposed the effects of search query design (original vs. ICB) using sentiment analysis. Together, the findings suggest that how a query is written can affect measures of precision, engagement, sentiment, and recall. Tests of precision are recommended to mitigate the effects of variance in retrieval relevance.

Does haze-related sentiment affect income inequality in China?

We investigate the effect of haze-related sentiment on urban–rural income inequality of prefecture-level cities in China. Income inequality is captured by the Theil index, the ratio and the difference between urban–rural income. We use GPT-3.5 models to conduct the sentiment analysis on the haze-related posts retrieved from the Weibo platform from 2018 to 2021. The generated emotion score and the number of posts are used to construct the haze-related sentiment measure. We find that haze-related unhappiness exacerbates urban–rural income inequality. The net population flow is investigated as the mechanism, indicating that negative sentiment reduces the net flows of both the floating population and resident population and consequently increases income inequality. The heterogeneity analysis finds that the effects of haze-related unhappiness on income inequality are smaller in cities with more hospital beds, doctors, and green spaces. Furthermore, the effects of the sentiment measure on income inequality vary significantly across cities in different geographic locations.

Trains, trade, and transformation: A spatial Rogowski theory of America's 19th‐century protectionism

AbstractWe study the effect of expanding trade on societal coalitions through its impact on development. We combine a majoritarian political model with a spatial model of trade to argue that trade‐induced economic change—by bringing new workers to locations closer to world markets—can lead to losses rather than gains in political power for the factors of production advantaged by increased trade. We study how this phenomenon explains rising protectionism in the United States from 1880 to 1900. Using county‐level changes in transportation costs induced by railroad expansion, our estimates indicate that falling costs increased population and farm values but reduced the proportion of farmers. Reduced transportation costs caused a reduction in vote shares for the Democratic Party, which favored liberal trade policies, and an increase in an original newspaper‐based measure of protectionist sentiment. Expanding trade alters not only political interests but also the geographic distribution of those interests.

Does time-varying risk aversion sentiment matter in the connectedness among Sub-Saharan African bond markets?

Several bond markets in sub-Saharan Africa (SSA) are defaulting due to hiking spreads amid the stressed states introduced by the COVID-19 pandemic and the geopolitical risk tensions from the Russia-Ukraine conflict. Are there controllable factors that drive these markets? We investigate the dynamic connection shared by SSA bond markets and assess the role of investor sentiment measures, focusing on the risk aversion sentiment of international investors. Our results, across different trading horizons, are expected to aid in the formulation of policies for regulating and developing bond markets of emerging economies, particularly SSA. In terms of both return and volatility of SSA bonds, we find risk aversion sentiment an important transmitter of spillover for all investment horizons.

The individual motivation of investors' gambling mentality: Empirical evidence from gambling sentiment and lottery‐type stocks

AbstractWe construct a measure of public gambling sentiment using the Baidu index and empirically examine the influence of temporal fluctuations in gambling sentiment on the performance of lottery‐type stocks. The empirical findings reveal that heightened gambling sentiment correlates with increased selling volume, short‐term order imbalances and liquidity in the lottery‐type stock market, while diminishing abnormal returns, including the initial returns of IPOs. We attribute this to investors' gambling mentality stemming from the accumulation of wealth rather than the pursuit of excitement, as they view lotteries and lottery‐related stocks as interchangeable investment options.

Equivalent sentiment measures for cross-language analysis of corporate communications

This paper presents a technique for sentiment measurement in many languages. The method allows researchers to efficiently analyze corporate documents, management reports, and financial statements using python. When the texts are written in many languages, the method extracts equivalent cross-linguistic sentiment features that can be used for statistical analysis or machine learning. We use Open Multilingual WordNet, a large lexicon organizing words into semantic groups, as the knowledge base about word equivalence in more than 200 languages. We experiment with a parallel English-French corpus and find that our senitment measures across the two languages are comparable. The method produces a consistent classification of positive and negative texts in two languages, and sentiment measure values correlate. The paper provides a detailed account of the method and python code, So that it can be applied to other languages, text mining, quantitative communication studies, and management research.•Method to create equivalent sentiment measures in multiple languages•Based on established lexicons and WordNet•Validated for English and French

Bank sentiment and liquidity hoarding

AbstractWe analyze how bank sentiment affects bank liquidity hoarding, distinguishing unexplained beliefs of bank managers from fundamental‐based beliefs. We build a bank management sentiment measure from textual analysis of 10‐Ks and utilize a comprehensive bank liquidity hoarding measure. We find that negative bank sentiment increases liquidity hoarding not warranted by a bank's fundamental conditions or external circumstances. Further analysis confirms that our findings reflect bank volition rather than being driven solely by borrowers or depositors. We address endogeneity concerns using exogenous weather conditions as instruments. Overall, our findings suggest that bank sentiment can influence how much liquidity banks provide to the economy and financial system.

Nowcasting Euro area GDP with news sentiment: A tale of two crises

SummaryThis paper shows that newspaper articles contain signals that can materially improve real‐time nowcasts of real GDP growth for the Euro area. Using articles from 15 popular European newspapers, which are machine translated into English, we create sentiment metrics that update daily and assess their value for nowcasting, comparing with competitive and rigorous benchmarks. We find that newspaper text is especially helpful early in the quarter before other indicators are available. We also find that general‐purpose sentiment measures perform better than more economics‐focused ones in response to unanticipated events and nonlinear supervised models can help capture extreme movements in growth but require sufficient training data to be effective.