The Impact of Customer Online Satisfaction on Stock Returns: Evidence from the E-commerce Reviews in China

Zhi Su, Danni Wu, Zhenkun Zhou, Junran Wu, Libo YinarXiv 2023 · DOI · arXiv

TL;DR

By mining over 18 million product reviews from JD.com, researchers have found that negative customer sentiment is a potent leading indicator for the Chinese stock market. Specifically, increases in negative review text and one-star ratings significantly predict lower future stock returns and negative earnings surprises, particularly for growth-oriented companies and firms with "opaque" accounting practices.

What Problem Does This Paper Address?

Traditional financial analysis relies heavily on "hard" data: quarterly reports, analyst forecasts, and regulatory filings. However, this information is often backward-looking and, in some markets, susceptible to management manipulation.

The authors address the gap in "alternative data" research within China’s A-share market. They seek to determine if the "wisdom of the crowd"—specifically the real-time, high-frequency feedback found on e-commerce platforms—contains novel information about a firm's fundamentals that isn't yet priced into the stock. The core question is: Can the collective grumbling of millions of online shoppers actually help investors predict a company's financial downfall before it hits the balance sheet?

Methods

The researchers employed a multi-step analytical pipeline:

  1. Sentiment Extraction: Using a tendentious lexicon, they quantified sentiment from review text to create the Customer Negative Sentiment Tendency (CNST) and Customer Positive Sentiment Tendency (CPST).
  2. Rating Analysis: They tracked the delta in one-star (OST) and five-star (FST) review ratios.
  3. Econometric Modeling: To handle the "chicken-and-egg" problem of endogeneity (e.g., does bad stock performance lead to bad reviews, or vice versa?), the authors used both static panel regressions with fixed effects and dynamic panel models using GMM (Generalized Method of Moments) estimators.
  4. Predictive Analysis: They tested whether these sentiment metrics could predict not just returns, but also "surprises" in revenue, earnings, and cash flow shocks using the HVZ and VOL models.

Datasets

  • Primary Dataset: A massive scrape of 18,008,415 customer reviews from JD.com, spanning 2008 to 2017.
  • Scope: The data covers 106 public firms and 164,715 unique products across four consumer-centric sectors: Home Appliances, Garments, Food, and Wine.
  • Financial Data: Stock returns and firm characteristics (Size, ROA, Book-to-Market, etc.) were sourced from the CSMAR (China Stock Market and Accounting Research) database.

Key Insights and Ideas

  • The Asymmetry of Sentiment: Interestingly, while negative sentiment (CNST) and one-star reviews (OST) were strong predictors of lower future returns, positive sentiment (CPST) had almost no predictive power. This suggests that markets are more sensitive to "red flags" from consumers than "gold stars."
  • The "Truth-Teller" Effect: The predictive power of customer reviews was significantly stronger for firms with low accounting transparency (those prone to earnings smoothing or aggressiveness). When official financial statements are opaque, the "boots-on-the-ground" data from e-commerce becomes a vital source of truth.
  • Growth vs. Value: Growth companies—whose valuations are tied to future expectations—are much more sensitive to online satisfaction trends than established "value" companies.
  • Sentiment Cycles: The impact of negative reviews is more pronounced during periods of high market sentiment. When the market is optimistic, negative consumer feedback acts as a sobering reality check that eventually forces a price correction.

Key Results

  • Return Predictability: Both CNST and OST are negatively associated with subsequent stock returns. A one standard deviation increase in CNST corresponds to a statistically significant drop in average weekly returns.
  • Fundamental Shocks: CNST successfully predicted negative revenue surprises, earnings surprises, and cash flow shocks. This confirms that online reviews aren't just "noise"; they reflect genuine changes in a company's ability to generate cash.
  • Economic Magnitude: In the static model, the coefficient for CNST was -0.0008 (t-stat of -2.7093), and for OST, it was -0.0072 (t-stat of -2.1665), indicating a robust and statistically significant negative impact on expected returns.

Limitations and Future Work

The authors acknowledge several areas for further exploration:

  • Sector Concentration: The study focused on four specific consumer industries. It remains to be seen how well this model translates to B2B industries or tech sectors where product reviews are less centralized.
  • Platform Bias: The data is exclusive to JD.com. Integrating data from other giants like Tmall or Pinduoduo could provide a more holistic view of the Chinese consumer landscape.
  • Lexicon Limitations: While the tendentious lexicon is effective, future work could utilize more advanced Deep Learning/NLP techniques (like BERT or GPT-based sentiment analysis) to capture nuance, sarcasm, or context-specific complaints that simple word counts might miss.

Summary updated

Abstract

This paper investigates the significance of consumer opinions in relation to value in China's A-share market. By analyzing a large dataset comprising over 18 million product reviews by customers on JD.com, we demonstrate that sentiments expressed in consumer reviews can influence stock returns, indicating that consumer opinions contain valuable information that can impact the stock market. Our findings show that Customer Negative Sentiment Tendency (CNST) and One-Star Tendency (OST) have a negative effect on expected stock returns, even after controlling for firm characteristics such as market risk, illiquidity, idiosyncratic volatility, and asset growth. Further analysis reveals that the predictive power of CNST is stronger in firms with high sentiment conditions, growth companies, and firms with lower accounting transparency. We also find that CNST negatively predicts revenue surprises, earnings surprises, and cash flow shocks. These results suggest that online satisfaction derived from big data analysis of customer reviews contains novel information about firms' fundamentals.
BibTeX
@misc{su2023impact,
  author = {Zhi Su and Danni Wu and Zhenkun Zhou and Junran Wu and Libo Yin},
  title = {{The Impact of Customer Online Satisfaction on Stock Returns: Evidence from the E-commerce Reviews in China}},
  year = {2023},
  eprint = {2306.12119},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2306.12119},
}
Figure