Digitization and Pre-Purchase Information
The Causal and Welfare Impacts of Reviews and Crowd Ratings
Summary
This paper aims to answer the question: Do consumers treat crowd-based reviews differently from professional reviews? In simple terms, the answer is yes.
Several related questions arise from this:
1. Do crowd-based and professional reviews influence consumers' purchasing decisions differently?
2. What are the overall benefits resulting from the introduction of crowd-based reviews?
The authors of this study focus on books sold on Amazon and examine the impact of reviews from two sources: the New York Times (NYT) and Amazon itself. They find that both types of reviews have positive effects on book sales on Amazon. However, the effects are not identical.
Reviews from the New York Times positively affect demand because they enhance the book's popularity. On the other hand, Amazon reviews serve two purposes. Firstly, they contribute to the book's popularity by improving its search rankings. Secondly, Amazon reviews allow consumers to gather more information about the product after conducting a search. The research reveals that Amazon reviews have a smaller, yet still positive, effect when books are relatively unknown. This suggests that Amazon reviews offer an additional dimension of information compared to NYT reviews, making them inherently different.
Both NYT and Amazon reviews assist consumers in making better book choices, thus improving overall welfare. However, the welfare effect resulting from Amazon reviews is ten times greater. This is mainly due to the vast number of products covered by Amazon reviews, whereas the NYT only reviews fewer than 2,000 books per year.
The bits and pieces
This paper studies how information can affect consumer purchase decision and hence welfare on the book market on Amazon. By purchasing a series of high-frequency datasets, the authors merge information about Amazon sales rank, price, ratings, and book reviews from prominent newspapers such as the New York Times. They also purchase data on book sales from Nielsen. They study 3 major English markets, Amazon USA, UK and Canada. However, we (and they) focus on the results as they relate to the USA.
They find that there exists a positive causal relationship between price, ratings, number of ratings and being reviewed in a major newspaper, and the rank of a product on Amazon. This relationship is particularly strong when a book is recommended by the New York Times. Fixing power law structure on book sales, they compute the elasticity of demand as it pertains to each variable.
They find that the elasticity of demand of reviews from newspapers fall off drastically in the number of days since the publication of the review. This indicates that newspaper reviews create an immediate shock to demand, but its effects wane very quickly. They show that on average, raters who have rated books reviewed by the NYT and other newspapers rate books that have not been reviewed by the news outlets lower. Arguing that they enjoy critic reviewed books more than other books.
Studying the elasticity of Amazon stars at different percentiles of sales rank, the authors show that at the higher percentiles, Amazon stars have a larger demand elasticity than at lower stars. This can account for the popularity of products with higher stars. But also highlights how products with less popularity can benefit from having ratings. Suggesting that ratings on the long tail of products with few ratings still have an influence on demand.
Imposing a structural model of demand, the authors study welfare arising from ratings. The paper estimates that ratings and reviews may account for an increase in 50 million dollars on spending on books in the USA. However, gains in consumer surplus on Amazon alone amounts to 3million from newspaper reviews, and 35 million from Amazon ratings. It is argued that the large difference is due to the absence of newspaper reviews in certain categories. Suggesting that the broadness of Amazon reviews allows consumers to make better decisions more often, driving welfare gains.
As interesting results that is not explored further, the paper also highlights how the effect of New York Times reviews on book sales has increased over the last 15 years. They also provide some evidence that the effect of rating becomes larger when the term is interacted with reviews. Suggesting that ratings may play a larger role on the point of sales while reviews play a more important role in search. More research would be interesting on these two topics.
The bit no one wants to see...
The paper focuses on Amazon star ratings from 2018, which is an aspect that many academics have moved past in favor of analyzing reviews. Incorporating Amazon reviews into the analysis would have been beneficial, as they provide additional insights beyond just the rating. The authors suggest that Amazon reviews could help identify the demand expansion effect and the impact of gaining extra information from viewing a product page.
The study quantifies that the welfare effect resulting from Amazon ratings is approximately 10 times greater than that of New York Times (NYT) reviews. However, it is reasonable to question this figure and propose a more conservative estimate of less than half that. While Amazon holds a 50% share of book sales in the USA, the other 50% comes from smaller outlets. These smaller stores are more likely to carry books reviewed by the NYT, benefiting from the demand expansion effect. Hence, learning about the quality of a popular book from an Amazon review is unlikely to have any effect here. Moreover, as such stores are less likely to stock unknown books, the spillover from Amazon reviews/ratings of less popular books to physical stores is minimal. Therefore, the welfare gain found online may not translate to offline stores.
It's important to note that using Amazon data for welfare analysis has its limitations. Amazon does not disclose actual sales numbers, so researchers often rely on sales rank information. While this can be helpful, factors like stock availability, review sentiment, returns, and other website interactions may influence relative sales ranks in ways we may not fully understand. While this is not a fair criticism of any paper that studies welfare on Amazon, it is a caveat that should be considered when reading papers them.