Research question
When are negative reviews more likely to be considered helpful? The study examines rating contrast, review content, and product category as three explanations for helpfulness judgments.
Data and methods
Using Amazon product-review data, the analysis models the logarithm of helpful votes. The workflow covers data cleaning, variable construction, text-topic classification, descriptive statistics, correlations, fixed-effects regressions, interaction models, heterogeneity analysis, and robustness checks.
Main findings
Negative reviews receive stronger helpfulness recognition when they contrast with a product's otherwise high average rating. Content that diagnoses risk—such as safety effects, service and logistics failures, authenticity, or value concerns—also tends to be perceived as more useful.
Boundaries
The findings explain information judgments in a public review environment. Helpful-vote behavior cannot be treated as a direct measure of purchase conversion or as causal evidence about consumer choice.