Julien Lefebvre

This paper of Chevalier and Mayzlin aims to show the effect of word of mouth on sales. As there is no existing literature on the outcome of online user reviews, authors tried to bring evidences about that effects, through the analysis of sales and reviews of two websites. They have chosen the main two on-line libraries in USA: Amazon.com and BarnesandNoble.com.

Authors show the limits of the investment of reviewing incentives for websites. First, why would people review on products instead of free-riding? Second, why websites would provide their own reviewing system whereas they can free ride with existing others? Third, what is the credibility of reviews because everyone can post his own following his proper interest? Last, reviews may not profit on sales.

Nevertheless, they focus on the differences between the two sites’ sales of books. They examine the relationship between the customer reviews at each site and firm sales by the analysis of book market share depending on the star rating. For instance they want to know for a well-reviewed and well liked book if a negative review on a website will drop the sales. In order to make the ideal experience books should have the same price and the same shipping conditions.

Authors generate a sample of books with two methods. First, a random sample of books selected from a catalog “Global Books in Print” that were published in 1998-2002. Second, books that appeared in the Publisher’s Weekly bestseller lists from 1991 to 2002. They removed books which have not been sold enough and does not appear in the sales rank of Amazon above the 650 000 place. The final sample contains 2394 observations, 1093 of which have reviews posted at both sites. They found that most of the books have a promised delivery of 24 hours (96% at Amazon and 88% at BN.com). They also found that BN.com prices are significantly higher and Amazon has more reviews than BN.com.

They demonstrate that there is no major preference differences between both sites, even if BN.Com customers tend to buy more non-fiction books than Amazon’s ones. They showed similarity regarding star rating across different categories for example juvenile fiction is the highest rated categories on both websites.

Then we arrive to the core of this article. What is the effect of reviews on sales? They analyze the differences of logarithm of ranking depending on the price, the review and the shipping. First they make a regression without the review variable. They obtain a price’s coefficient in the regression for both sites that implies when price raise sales drop. However, this effect seems to be stronger at BN.com.

Afterward they make a regression that included the variable on reviews. As Chevalier and Goolsbee already prove sales ranking are lower on both sites while numbers of reviews are high. They estimate for a book which sales ranking is 500 and haven’t got any review, a posting of 3 reviews on Amazon will jump rank to 327. One might consider that the number of reviews could be the effect of sales. But reviews are much older than sales only 2% of reviews were posted during the previous two months whereas sales rank is actual (week or month).

But this effect is an average effect; consequently the star ranking should be included. As expected for both sites, the coefficients for the average star value suggest that sales improve when books are rated more highly, but the effect is statistically insignificant for BN.com. For example, a book with a sales ranking of 500 which has four 5-stars reviews. If you turn a review into a 1-star review the effect on the ranking will be a lift to the 603 rang on Amazon whereas the BN.com ranking would not be affect.

A 1-star review has far more effect on sales than 5-stars reviews. This is explained by the credibility of the review. Any authors can easily post a large number of positive reviews cheaply whereas negative ones are posted by unsatisfied people.

The differences of preference between customers of both sites affect both sales and ranking. These differences are subtle because the preferences on categories are identical. About the longer of reviews they found that reviews are longer for 4-star than 5-star because reviewer also have to talk about the negative aspect. However, the longer of sales does not stimulate sales.

To sum up, authors show reviews tend to be very positive at both sites, that they are more detailed on Amazon. Thanks to their regression they estimate that the relative market share of a book across the sites is related to the numbers of reviews and the star ranking. Therefore customer word-of-mouth has a causal impact on consumer purchasing behavior at two Internet retail sites.