Pith. sign in

REVIEW 5 major objections 4 minor 46 references

Quantitative Analysis of Media Bias and Stock Price Dynamics: The 2020 Shock

T0 review · 5 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read After the 2020 shock, neither media stance nor firm returns changed level once common market moves and firm differences are removed, and no market-wide lead-lag relationship survives.

desk verdict A genuinely useful firm-level framework undone by a textbook collinearity error in the two headline regressions. read the letter →

arxiv 2608.05899 v1 pith:QGG5WPA5 submitted 2026-08-06 cs.CE cs.SI

Shivansh Verma , Soham Tulsyan , Sashwat Dhanuka , Anirban Sen This is my paper · ORCID
classification cs.CEcs.SI MSC 62M1062P2091B84
keywords mediasentimentfirm-levelstanceCOVID-19shockstructuralbreakspanelVARGrangercausalityefficientmarkethypothesisnewsandstockreturns
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper asks whether the COVID-19 shock changed how financial news and stock prices interact, using 90,579 materially relevant headlines for 26 large U.S. firms from 2015 to 2025. It tries to establish that after removing firm-specific differences and common daily market movements, neither the tone of a firm's coverage nor its returns show a persistent level shift around 2020. It further argues that dynamic links between stance and returns exist only for a subset of firms, typically after each firm's own estimated structural break, and vanish when firms are pooled. This matters because most prior work measures sentiment at the market level, so it cannot tell a genuine market-wide media effect from the combined effect of a few firm-level stories. If the paper is right, studies of news and markets should move from aggregate indices to firm-specific measurement and data-dated breaks.

What carries the argument

The central object is a firm-day stance measure built from target-dependent sentiment scores: for each headline about firm i, NewsMTSC gives probabilities p_+, p_-, p_0 toward that firm, collapsed into the signed score p_+ - p_- and averaged over the day's retained headlines. Two-way fixed-effects panel regressions (firm effects plus day or month effects) carry the level-shift tests, isolating within-firm change from common market movements. Vector autoregressions on weekly differenced stance and returns, with BIC lag selection, Arellano-Bover Helmert transformation for the pooled panel, and Bai-Perron data-driven break dating, carry the dynamic tests. The machinery separates firm-specific dynamics from common shocks and lets each firm's break date come from the data rather than from the calendar.

What would settle it

Validate the stance measure by having financial annotators label a random sample of the 90,579 retained headlines for tone toward the named firm and compare agreement with NewsMTSC scores; if agreement is low or errors correlate with firm, year, or break timing, the level-shift and Granger results rest on mismeasured stance. Alternatively, rerun RQ3 with a finance-domain-adapted sentiment model and check whether market-wide Granger causality appears where the paper reports none.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is a pair of nulls plus a localization result. With firm and day fixed effects and a coverage-volume control, the post-2020 coefficient on daily stance is +0.011 with p = 0.71; with firm and month effects plus S&P 500 and VIX controls, the post-2020 coefficient on daily log returns is -0.0001 with p = 0.92. In firm-by-firm vector autoregressions, only Uber shows stance forecasting returns decisively and only Goldman Sachs shows returns forecasting stance over the full sample; splitting at Bai-Perron breaks brings out post-break channels for a handful of firms. Pooling all firms in a Helmert-transformed panel VAR with market-wide controls yields no significant Granger causality in either direction in any subsample. The paper reads this as: the 2020 shock left no common mark on tone or returns, and the news-return link, where real, belongs to individual firms and their own break dates.

Load-bearing premise

The load-bearing premise is that NewsMTSC's political-news sentiment scores measure financial tone correctly for the 26 firms; the model was applied without retraining or validation on financial headlines, and if it mis-scores financial language, the null results and firm-level findings could be artifacts of measurement error.

Editorial extensions

If this is right

  • Aggregate sentiment studies may be detecting effects driven by a minority of firms rather than by a market-wide news-to-price mechanism.
  • The absence of a stance level shift suggests the pandemic did not systematically bend press coverage for or against large firms once common news-cycle effects are absorbed.
  • Data-dated structural breaks, rather than calendar-selected dates such as March 2020, are needed to uncover firm-level regime changes in news-return dynamics.
  • The null return shift is consistent with efficient pricing of a market-wide event: a common repricing occurs, but no residual firm-level step remains after controls.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A natural next test is to re-run the analysis with a finance-domain sentiment model; if strong firm-level or aggregate predictability emerges, the paper's nulls may partly reflect measurement error from applying a political-news stance model to financial headlines.
  • The paper's design implies that event-study analyses should estimate each firm's own break date; imposing March 2020 would have missed Wells Fargo's 2019 stance break and 2021 return break.
  • Densely covered firms could be pushed to daily or event-time frequency to see whether weekly aggregation hides a fast news-to-price channel.
  • A distributional summary of daily tone, rather than a signed mean, could reveal changes in coverage disagreement that the level tests cannot see; the paper itself flags this extension.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 4 minor

Summary. The paper investigates whether the COVID-19 shock changed the level or dynamic relationship between firm-specific media stance and stock returns for 26 large US firms, using 6.28 million headlines filtered to 90,579 relevant ones. The authors estimate panel regressions with firm and time fixed effects to test for post-2020 level shifts (RQ1 and RQ2) and use firm-level and panel VARs with Bai-Perron structural breaks to test Granger causality (RQ3). They conclude that neither stance nor returns exhibit a persistent level shift, but that dynamic relationships emerge for a subset of firms around their own breaks, with no market-wide lead-lag relationship.

Significance. If the results were credible, they would offer a useful firm-level complement to aggregate sentiment studies and a caution against market-wide generalizations. The dataset construction is relatively transparent, and the authors make an effort to control for common shocks and to date breaks empirically. However, the central econometric identification is flawed, and the headline null results are not identified by the specified regressions. The contribution is therefore conditional on a fix to the level-shift tests and on addressing the measurement and multiple-testing issues.

major comments (5)
  1. [IV-B, Eq. (2), Table III(a)] The regressor Post_t = 1[t >= 2020-01-01] is a deterministic function of time only and is perfectly collinear with the set of day fixed effects δ_day_t. After the within transformation that removes firm and day means, the demeaned Post variable is identically zero, so β is unidentified. The reported coefficient +0.011 (p=0.71) is an arbitrary normalization (e.g., whichever day dummy is dropped) and cannot be interpreted as evidence against a level shift. This undermines the abstract and conclusion claim that media stance did not shift after 2020.
  2. [IV-C, Eq. (3), Table III(b)] The same identification failure occurs in the returns regression: Post_t is constant within each calendar month and is absorbed by the monthly fixed effects δ_month_t. The reported β = −0.0001 (p=0.92) is arbitrary and does not test whether firm-level returns shifted after 2020. The conclusion that firm returns show no level break is unsupported by the regression as specified.
  3. [V-C, Table IV] The firm-level Granger tests are run for 26 firms in two directions and on three samples (full, pre-break, post-break), which is at least 156 tests. The paper reports only a handful of p-values and does not apply any multiple-testing correction. At the 5% level one would expect about 8 significant results by chance even if no relationship exists, so the evidence for 'a subset of firms' is weak without a full reporting of all tests or an FDR control.
  4. [IV-D, V-C] The Bai-Perron procedure estimates a break in the mean of each series on the very same data that are then split into pre- and post-break segments for the Granger tests. Because the break is selected to maximize the fit in the dependent variable and its uncertainty is ignored, the post-break p-values are likely to understate the true variability. The paper should report the break dates for all 26 firms, justify using the mean break rather than a break in the Granger coefficients, and assess sensitivity to break-date uncertainty.
  5. [II-B, IV-A] The stance scores come from NewsMTSC, a target-dependent sentiment model trained on political news, and the paper gives no evidence that its scores are valid for financial headlines. The related-work section itself emphasizes that finance-adapted models outperform general-purpose ones (e.g., references [17]–[19]). Without a validation study (e.g., a labeled financial headline sample or comparison against a financial sentiment benchmark), systematic measurement error in the stance variable could both mask real effects and create spurious ones, threatening all three research questions.
minor comments (4)
  1. [II-C] The sentence 'The bivariate design carries only stance and sector shocks, sit outside it' is ungrammatical and unclear; it should likely read 'The bivariate design carries only stance and returns; other drivers, such as sector shocks, sit outside it.'
  2. [IV-D] The paper states that the stance index is 'stationary for some firms and integrated or break-driven for the rest,' but it does not provide the ADF/KPSS results or a table of the Bai-Perron break dates. Reporting this information would make the analysis reproducible.
  3. [V-C, Table IV] Table IV reports only the firms with significant Granger results, so the reader cannot judge the overall false-positive rate. A full table or a summary of the distribution of p-values is needed.
  4. [IV-A] The relevance classifier's operating threshold is chosen to favor precision, but the paper does not discuss how the threshold choice affects the stance time series, for example through a sensitivity analysis with alternative thresholds.

Circularity Check

1 steps flagged · score 6.0 of 10

RQ1 and RQ2 level-shift nulls are forced by construction: Post_t is a linear combination of the day/month fixed effects, so the reported coefficients are unidentified.

  1. self definitional [Section IV-B, Eq. (2); Section IV-C, Eq. (3); results in Section V-A and V-B]
    "biasit = α_i + δ_day_t + β Post_t + γ vol_it + ε_it (2). Here Post_t = 1[t≥2020-01-01] switches on for every day from January 2020 onward, and its coefficient β is the quantity of interest; α_i are the firm effects and δ_day_t the daily effects. returnit = α_i + δ_month_t + β Post_t + γ_1 sp500 ret_t + γ_2 vix_t + ε_it (3), with Post_t as before, α_i the firm effects, δ_month_t the monthly effects."

    Post_t varies only over time and is identical across firms. With a full set of day dummies in Eq. (2), Post_t equals the sum of the post-2020 day dummies, a perfect linear combination of the daily effects. After the within-transformation, the demeaned Post regressor is identically zero, so β is unidentified; the reported +0.011 is an arbitrary artifact of the dropped-dummy normalization, not a data estimate. The same holds for Eq. (3): Post_t is constant within each calendar month and is absorbed by the monthly fixed effects. The headline conclusion that neither stance nor returns exhibit a persistent level shift is therefore not an empirical test but a consequence of the specification: any common time shift is removed by the fixed effects before β can be estimated.

full rationale

The paper is an empirical panel study with no self-citation chain and no fitted parameter renamed as a prediction. The relevance classifier is trained on human labels, the stance model is an external tool (NewsMTSC), and the RQ3 VAR/Granger analysis is a standard predictive exercise. The Bai-Perron breaks are estimated on the mean of each series, not on the Granger coefficients, so splitting the sample at those breaks does not by itself force the dynamic nulls. The load-bearing circularity is confined to RQ1 and RQ2: the post-shock indicator is collinear with the time fixed effects, making the two headline nulls artifacts of the specification. The paper's own Section VI-B admits that removing the common component 'stays silent on the large aggregate repricing,' which is the same construction. Because the abstract and conclusion lead with these two nulls, the circularity is partial but substantial; RQ3 remains independent, so a score of 6 is appropriate.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new theoretical entities. Its contributions are empirical, and the main explicit choices are the classifier threshold and training weights. The key assumptions concern the validity of the sentiment model and the transferability of classifier performance from a small test set to the full corpus.

free parameters (2)
  • Relevance classifier decision threshold = 0.80
    Chosen by hand to favor precision for full-corpus inference; directly controls which headlines enter the stance series and thus all downstream estimates.
  • Human-label weight in classifier training = 5x
    Hyperparameter weighting human annotations relative to synthetic ones; affects classifier calibration and the resulting corpus.
assumptions (5)
  • domain assumption Stance of a firm's coverage is well approximated by the mean of headline-level target-dependent sentiment scores (Eq. 1).
    The daily stance measure collapses opposing headlines into a single mean, assuming the center summarizes tone.
  • domain assumption NewsMTSC, trained on political news, produces valid target-dependent sentiment scores for financial headlines.
    No financial-domain evaluation is provided; the paper's own review argues domain adaptation matters.
  • domain assumption The relevance classifier's precision and recall on the test set (n=187) extend to the full corpus distribution.
    The classifier was trained on 750 human labels plus synthetic data, and the operating threshold was chosen in part after inspecting test-set performance curves.
  • domain assumption Bai-Perron single-break estimates provide valid sample splits for downstream Granger tests.
    Break dates are estimated on the same series used for testing, and the paper does not account for this specification search or multiple testing across firms.
  • standard math Standard OLS, VAR, and Granger-causality asymptotic theory applies at these sample sizes.
    The paper uses HC1, HAC, and Newey-West standard errors, which require large samples and stationarity; some firm segments have short post-break series.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Quantitative Analysis of Media Bias and Stock Price Dynamics: The 2020 Shock." pith.science (2026). https://pith.science/paper/QGG5WPA5

@misc{pith2026260805899,
  author       = {Pith},
  title        = {Pith review of: Quantitative Analysis of Media Bias and Stock Price Dynamics: The 2020 Shock},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QGG5WPA5}},
  note         = {Machine review of arXiv:2608.05899}
}
read the original abstract

Whether financial news influences stock prices or simply reflects information already incorporated into them remains an open question in financial economics. The COVID-19 pandemic provides an opportunity to revisit this question, as it disrupted both news coverage and financial markets on an unprecedented scale. Existing studies have largely approached the problem through aggregate sentiment measures, leaving it unclear whether the observed relationships also hold at the level of individual firms. We study this question using 6.28 million news headlines covering 26 large United States firms between 2015 and 2025. After filtering the corpus to retain materially relevant firm-specific coverage, we construct daily stance measures and examine how their relationship with stock returns changed around the 2020 shock using panel regressions and vector autoregressions with data-driven structural breaks. Our findings indicate that the relationship between financial news and equity markets is more nuanced than aggregate analyses alone suggest. While we find little evidence of a persistent market-wide change in media stance or stock returns following the pandemic, dynamic relationships emerge for a subset of firms around their own structural breaks. Taken together, these results suggest that understanding media-market interactions requires firm specific analysis and provide a framework for studying how news and prices interact under changing market conditions.

Figures

Figures reproduced from arXiv: 2608.05899 by the authors.

Figure 1
Figure 1. Yearly headline coverage for each candidate firm, shaded by [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. The relevance-classifier pipeline. A raw S&P 500 news headline is [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 5
Figure 5. Visa weekly log returns, 2015–2025. The March 2020 crash (dashed [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figures from the paper (1 more)
Figure 6
Figure 6. Figure 6: Data-driven structural breaks for Wells Fargo. The Bai–Perron break in [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 41 canonical work pages

  1. [17]

    FinBERT: Financial sentiment analysis with pre-trained language models,

    D. Araci, “FinBERT: Financial sentiment analysis with pre-trained language models,” arXiv preprint arXiv:1908.10063, 2019

    arXiv 1908

  2. [19]

    Advanced financial sentiment analysis using FinBERT to explore sen- timent dynamics,

    S. Baghavathi Priya, M. Kumar, J. D. Nitheesh Prakash, and N. Krithika, “Advanced financial sentiment analysis using FinBERT to explore sen- timent dynamics,” inProceedings of the 3rd International Conference on Intelligent Data Communication Technologies and Internet of Things (IDCIoT). IEEE, 2025, pp. 889–897

    work page 2025

  3. [1]

    Efficient capital markets: A review of theory and empirical work,

    E. F. Fama, “Efficient capital markets: A review of theory and empirical work,”The Journal of Finance, vol. 25, no. 2, pp. 383–417, 1970

    1970

  4. [2]

    NewsMTSC: A dataset for (multi-)target- dependent sentiment classification in political news articles,

    F. Hamborg and K. Donnay, “NewsMTSC: A dataset for (multi-)target- dependent sentiment classification in political news articles,” inProceed- ings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume, 2021, pp. 1663–1675

    work page 2021

  5. [3]

    Machine learning sen- timent analysis, COVID-19 news and stock market reactions,

    M. Costola, O. Hinz, M. Nofer, and L. Pelizzon, “Machine learning sen- timent analysis, COVID-19 news and stock market reactions,”Research in International Business and Finance, vol. 64, p. 101881, 2023

    work page 2023

  6. [4]

    Financial market sentiment and stock return during the COVID-19 pandemic,

    C. Bai, Y . Duan, X. Fan, and S. Tang, “Financial market sentiment and stock return during the COVID-19 pandemic,”Finance Research Letters, vol. 54, p. 103709, 2023

    work page 2023

  7. [5]

    Feverish sentiment and global equity markets during the COVID-19 pandemic,

    T. L. D. Huynh, M. Foglia, M. A. Nasir, and E. Angelini, “Feverish sentiment and global equity markets during the COVID-19 pandemic,” Journal of Economic Behavior & Organization, vol. 188, pp. 1088–1108, 2021

    work page 2021

  8. [6]

    Constructing a positive sentiment index for COVID-19: Evidence from G20 stock markets,

    D. Anastasiou, A. Ballis, and K. Drakos, “Constructing a positive sentiment index for COVID-19: Evidence from G20 stock markets,” International Review of Financial Analysis, vol. 81, p. 102111, 2022

    work page 2022

Show all 46 references
  1. [7]

    Positive COVID-19 related sentiment, economic uncertainty and risk manage- ment implications,

    D. Anastasiou, A. Ballis, C. Kallandranis, and I. Vlassas, “Positive COVID-19 related sentiment, economic uncertainty and risk manage- ment implications,”Journal of Banking Regulation, vol. 27, no. 1, pp. 1–13, 2026

    2026

  2. [8]

    Economic news, social media sentiments, and stock returns: Which is a bigger driver?

    R. Verma and P. Verma, “Economic news, social media sentiments, and stock returns: Which is a bigger driver?”Journal of Risk and Financial Management, vol. 18, no. 1, p. 16, 2025

    2025

  3. [9]

    Does it really pay off for investors to consider information from social media?

    B. Eierle, S. Klamer, and M. Muck, “Does it really pay off for investors to consider information from social media?”International Review of Financial Analysis, vol. 81, p. 102074, 2022

    2022

  4. [10]

    News vs. social media: Sentiment impact on stock performance of big tech companies,

    H. Kim-Hahm, A. S. Abou-Zaid, and A. Mohd, “News vs. social media: Sentiment impact on stock performance of big tech companies,”Journal of Risk and Financial Management, vol. 18, no. 12, p. 660, 2025

    2025

  5. [11]

    Investor attention and reaction in COVID- 19 crisis: sentiment analysis in the Indian stock market,

    N. B. Sing and R. G. Singh, “Investor attention and reaction in COVID- 19 crisis: sentiment analysis in the Indian stock market,”Managerial Finance, vol. 49, no. 3, pp. 470–491, 2023

    2023

  6. [12]

    Does media sentiment affect stock prices? evidence from China’s STAR market,

    X. Dong, S. Xu, J. Liu, and F.-S. Tsai, “Does media sentiment affect stock prices? evidence from China’s STAR market,”Frontiers in Psy- chology, vol. 13, p. 1040171, 2022

    2022

  7. [13]

    The way digitalization is impacting international financial markets: Stock price synchronicity,

    C. Chen, M. M. Moeini Gharagozloo, L. Darougar, and L. Shi, “The way digitalization is impacting international financial markets: Stock price synchronicity,”International Finance, vol. 25, no. 3, pp. 396–415, 2022

    2022

  8. [14]

    Firm-level investor sentiment and corporate announcement returns,

    N. Mahmoudi, P. Docherty, and A. Melia, “Firm-level investor sentiment and corporate announcement returns,”Journal of Banking & Finance, vol. 144, p. 106586, 2022

    2022

  9. [15]

    A multi- level sentiment analysis framework for financial texts,

    Y . Liu, J. Wang, L. Long, X. Li, R. Ma, Y . Wu, and X. Chen, “A multi- level sentiment analysis framework for financial texts,” arXiv preprint arXiv:2504.02429, 2025

    2025 arXiv

  10. [16]

    Stock price prediction using FinBERT-enhanced sentiment with SHAP explainability and differential privacy,

    L. Ruan and H. Jiang, “Stock price prediction using FinBERT-enhanced sentiment with SHAP explainability and differential privacy,”Mathemat- ics, vol. 13, no. 17, p. 2747, 2025

    2025

  11. [18]

    FinBERT: A pre- trained financial language representation model for financial text min- ing,

    Z. Liu, D. Huang, K. Huang, Z. Li, and J. Zhao, “FinBERT: A pre- trained financial language representation model for financial text min- ing,” inProceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence (IJCAI-20), 2020, pp. 4513–4519

    2020

  12. [20]

    FinEntity: Entity- level sentiment classification for financial texts,

    Y . Tang, Y . Yang, A. Huang, A. Tam, and J. Tang, “FinEntity: Entity- level sentiment classification for financial texts,” inProceedings of the 2023 Conference on Empirical Methods in Natural Language Process- ing, Singapore, 2023, pp. 15 465–15 471

    2023

  13. [21]

    Entity-level sentiment analysis (ELSA): An exploratory task survey,

    E. Rønningstad, E. Velldal, and L. Øvrelid, “Entity-level sentiment analysis (ELSA): An exploratory task survey,” inProceedings of the 29th International Conference on Computational Linguistics, Gyeongju, Republic of Korea, 2022, pp. 6773–6783. XI S. VERMA, S. TULSY AN, S. DHA...

    2022

  14. [22]

    A multi-source entity-level sentiment corpus for the finan- cial domain: the FinLin corpus,

    T. Daudert, “A multi-source entity-level sentiment corpus for the finan- cial domain: the FinLin corpus,”Language Resources and Evaluation, vol. 56, pp. 333–356, 2022

    2022

  15. [23]

    Evaluating large language models for stance detection on financial targets from SEC filing reports and earnings call transcripts,

    N. Gyawali, D. Caragea, A. Vasenkov, and C. Caragea, “Evaluating large language models for stance detection on financial targets from SEC filing reports and earnings call transcripts,” arXiv preprint arXiv:2510.23464, 2025

    2025

  16. [24]

    Beyond correlation: Refutation-validated aspect-based sentiment analysis for explainable energy market returns,

    W. van der Heever, K. Ong, R. Satapathy, and E. Cambria, “Beyond correlation: Refutation-validated aspect-based sentiment analysis for explainable energy market returns,” arXiv preprint arXiv:2603.21473, 2026

    2026

  17. [25]

    Are ChatGPT and GPT-4 general-purpose solvers for financial text analytics? a study on several typical tasks,

    X. Li, S. Chan, X. Zhu, Y . Pei, Z. Ma, X. Liu, and S. Shah, “Are ChatGPT and GPT-4 general-purpose solvers for financial text analytics? a study on several typical tasks,” inProceedings of the 2023 Conference on Empirical Methods in Natural Language Processing: Industry Track...

    2023

  18. [26]

    Large language models in finance: what is financial sentiment?

    K. Kirtac and G. Germano, “Large language models in finance: what is financial sentiment?” arXiv preprint arXiv:2503.03612, 2025

    2025 arXiv

  19. [27]

    Reasoning or overthinking: Evaluating large language models on financial sentiment analysis,

    D. Vamvourellis and D. Mehta, “Reasoning or overthinking: Evaluating large language models on financial sentiment analysis,” inProceedings of the 6th ACM International Conference on AI in Finance (ICAIF ’25), 2025, pp. 299–307

    2025

  20. [28]

    Can AI read between the lines? benchmarking LLMs on financial nuance,

    D. Kubica, D. T. Gordon, N. Emura, D. Saini, and C. Goldenberg, “Can AI read between the lines? benchmarking LLMs on financial nuance,” arXiv preprint arXiv:2505.16090, 2025

    2025 arXiv

  21. [29]

    Prompt sentiment: The catalyst for LLM change,

    V . Gandhi and S. Gandhi, “Prompt sentiment: The catalyst for LLM change,” arXiv preprint arXiv:2503.13510, 2025

    2025 arXiv

  22. [30]

    Fin-Bias: Comprehensive evaluation for LLM decision-making under human bias in the finance domain,

    X. Hu and J. Zhao, “Fin-Bias: Comprehensive evaluation for LLM decision-making under human bias in the finance domain,” arXiv preprint arXiv:2605.09106, 2026

    2026 arXiv

  23. [31]

    Look-Ahead-Bench: a standardized benchmark of look-ahead bias in point-in-time LLMs for finance,

    M. Benhenda, “Look-Ahead-Bench: a standardized benchmark of look-ahead bias in point-in-time LLMs for finance,” arXiv preprint arXiv:2601.13770, 2026

    2026

  24. [32]

    Fake date tests: Can we trust in-sample accuracy of LLMs in macroeconomic forecasting?

    A. Eliseev and S. Seleznev, “Fake date tests: Can we trust in-sample accuracy of LLMs in macroeconomic forecasting?” arXiv preprint arXiv:2601.07992, 2026

    2026

  25. [33]

    Incorporating stock market signals for Twitter stance detection,

    C. Conforti, J. Berndt, M. T. Pilehvar, C. Giannitsarou, F. Toxvaerd, and N. Collier, “Incorporating stock market signals for Twitter stance detection,” inProceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Dublin, I...

    2022

  26. [34]

    Examining time-varying causality: investor sentiment and asset spreads across COVID and Ukraine war periods,

    V . F. Moutinho, R. H. Correia Domingues, G. Fantini, and M. Moraes, “Examining time-varying causality: investor sentiment and asset spreads across COVID and Ukraine war periods,”Applied Economics Letters, 2025

    2025

  27. [35]

    Investor sentiments and stock markets during the COVID-19 pandemic,

    E. Cevik, B. Kirci Altinkeski, E. I. Cevik, and S. Dibooglu, “Investor sentiments and stock markets during the COVID-19 pandemic,”Finan- cial Innovation, vol. 8, no. 1, p. 69, 2022

    2022

  28. [36]

    News and markets in the time of COVID-19,

    H. Mamaysky, “News and markets in the time of COVID-19,”Journal of Financial and Quantitative Analysis, vol. 59, no. 8, pp. 3564–3600, 2024

    2024

  29. [37]

    Firm level return–volatility analysis using dynamic panels,

    L. V . Smith and T. Yamagata, “Firm level return–volatility analysis using dynamic panels,”Journal of Empirical Finance, vol. 18, no. 5, pp. 847– 867, 2011

    2011

  30. [38]

    Structural breaks in online investor sentiment: A note on the nonstationarity of financial chatter,

    D. Ballinari and S. Behrendt, “Structural breaks in online investor sentiment: A note on the nonstationarity of financial chatter,”Finance Research Letters, vol. 35, p. 101479, 2020

    2020

  31. [39]

    Impact of COVID-19 on stock indices volatility: Long-memory persistence, structural breaks, or both?

    A. M. B. de Oliveira, A. Mandal, and G. J. Power, “Impact of COVID-19 on stock indices volatility: Long-memory persistence, structural breaks, or both?”Annals of Data Science, vol. 11, no. 2, pp. 619–646, 2024

    2024

  32. [40]

    Structural breaks in interactive effects panels and the stock market reaction to COVID-19,

    Y . Karavias, P. K. Narayan, and J. Westerlund, “Structural breaks in interactive effects panels and the stock market reaction to COVID-19,” Journal of Business & Economic Statistics, vol. 41, no. 3, pp. 653–666, 2023

    2023

  33. [41]

    Multiple structural breaks in interactive effects panel data models,

    J. Ditzen, Y . Karavias, and J. Westerlund, “Multiple structural breaks in interactive effects panel data models,”Journal of Applied Econometrics, vol. 40, no. 1, pp. 74–88, 2025

    2025

  34. [42]

    Testing and estimating structural breaks in time series and panel data in Stata,

    ——, “Testing and estimating structural breaks in time series and panel data in Stata,”The Stata Journal, vol. 25, no. 3, pp. 526–560, 2025

    2025

  35. [43]

    Estimation of panel group structure models with structural breaks in group memberships and coefficients,

    R. L. Lumsdaine, R. Okui, and W. Wang, “Estimation of panel group structure models with structural breaks in group memberships and coefficients,”Journal of Econometrics, vol. 233, no. 1, pp. 45–65, 2023

    2023

  36. [44]

    Discovering what mattered: Detecting unknown treatment as breaks in panel models,

    F. Pretis and M. Schwarz, “Discovering what mattered: Detecting unknown treatment as breaks in panel models,” SSRN Working Paper 4022745, 2026

    2026

  37. [45]

    Vector autoregressive-based Granger causality test in the presence of instabilities,

    B. Rossi and Y . Wang, “Vector autoregressive-based Granger causality test in the presence of instabilities,”The Stata Journal, vol. 19, no. 4, pp. 883–899, 2019

    2019

  38. [46]

    Disaster resilience and asset prices,

    M. Pagano, C. Wagner, and J. Zechner, “Disaster resilience and asset prices,”Journal of Financial Economics, vol. 150, no. 2, p. 103712, 2023

    2023

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.