Big Data and Text Analytics on 833K Glassdoor Reviews: Aspect-Level Employee Sentiment as a Short-Horizon Early-Warning Signal of Firm Distress
Abstract
We test whether the tone of free-text employee reviews, decomposed into specific human-resource (HR) aspects, anticipates near-term deterioration in a firm’s standing on the same review platform. Using 833,760 Glassdoor reviews of 332 large employers (2008–2021), we measure sentiment on five aspects—management, work–life balance, compensation, culture, and career growth—with a transparent, reproducible lexicon-based aspect sentiment procedure, and we validate this text signal against the platform’s own human-assigned sub-ratings (Pearson r between 0.42 and 0.52). In a firm–month panel with two-way (firm and calendar-month) fixed effects and firm-clustered standard errors, more negative management, culture, and career-growth sentiment in month t−1 is associated with a significantly higher probability of a firm-level rating decline in month t. The association is statistically robust but economically modest (within-R² ≈ 0.027), and—importantly—it is short-horizon: at a three-month lead the predictive content essentially vanishes (within-R² ≈ 0.000; joint F not significant). An event study around 703 distress onsets shows aspect sentiment dipping roughly one to two months before the rating trough, and Granger-precedence tests support a lead for a majority (14 of 25) of long-series firms. We deliberately frame the outcome as an internal platform signal rather than a hard external event (e.g., layoffs): the canonical external layoffs dataset does not overlap the review window in time, precluding external validation here. Results should therefore be read as evidence that aspect sentiment leads a soft, platform-internal distress proxy at a short horizon, not as a validated predictor of corporate restructuring.
Keywords
Downloads
References
- Blei, D.M., Ng, A.Y., Jordan, M.I. (2003). Latent Dirichlet Allocation. Journal of Machine Learning Research, 3, 993–1022. https://www.jmlr.org/papers/volume3/blei03a/blei03a.pdf
- Devlin, J., Chang, M.-W., Lee, K., Toutanova, K. (2019). BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. NAACL-HLT. https://aclanthology.org/N19-1423/
- Edmans, A. (2011). Does the stock market fully value intangibles? Employee satisfaction and equity prices. Journal of Financial Economics, 101(3), 621–640. https://doi.org/10.1016/j.jfineco.2011.03.021
- Granger, C.W.J. (1969). Investigating Causal Relations by Econometric Models and Cross-spectral Methods. Econometrica, 37(3), 424–438. https://www.jstor.org/stable/1912791
- Green, T.C., Huang, R., Wen, Q., Zhou, D. (2019). Crowdsourced employer reviews and stock returns. Journal of Financial Economics, 134(1), 236–251. https://doi.org/10.1016/j.jfineco.2019.03.012
- Grootendorst, M. (2022). BERTopic: Neural topic modeling with a class-based TF-IDF procedure. arXiv:2203.05794. https://arxiv.org/abs/2203.05794
- He, P., Liu, X., Gao, J., Chen, W. (2021). DeBERTa: Decoding-enhanced BERT with Disentangled Attention. ICLR. https://openreview.net/forum?id=XPZIaotutsD (arXiv: https://arxiv.org/abs/2006.03654)
- Hutto, C.J., Gilbert, E. (2014). VADER: A Parsimonious Rule-based Model for Sentiment Analysis of Social Media Text. ICWSM. https://ojs.aaai.org/index.php/ICWSM/article/view/14550
- Lee, D.D., Seung, H.S. (1999). Learning the parts of objects by non-negative matrix factorization. Nature, 401, 788–791. https://www.nature.com/articles/44565
- Liu, B. (2012). Sentiment Analysis and Opinion Mining. Morgan & Claypool. https://www.cs.uic.edu/~liub/FBS/SentimentAnalysis-and-OpinionMining.pdf
- Pang, B., Lee, L. (2008). Opinion Mining and Sentiment Analysis. Foundations and Trends in Information Retrieval, 2(1–2), 1–135. https://doi.org/10.1561/1500000011
- Petersen, M.A. (2009). Estimating Standard Errors in Finance Panel Data Sets: Comparing Approaches. Review of Financial Studies, 22(1), 435–480. https://doi.org/10.1093/rfs/hhn053
- Pontiki, M., Galanis, D., Pavlopoulos, J., et al. (2014). SemEval-2014 Task 4: Aspect Based Sentiment Analysis. Proceedings of SemEval-2014. https://aclanthology.org/S14-2004/
- Sull, D., Sull, C., Zweig, B. (2022). Toxic Culture Is Driving the Great Resignation. MIT Sloan Management Review, 63(2). https://sloanreview.mit.edu/article/toxic-culture-is-driving-the-great-resignation/