I've been thinking recently about Amazon's reviewing system. On the couple of times that I have reviewed something harshly, for example my review of The Shadow of the Wind in which I contradict a number of people, I've received a number of 'not helpful' votes. My problem with this rating system is the following - I think that liking something and finding something helpful are two distinct notions. But, I believe that people use the voting system to indicate dislike of something, regardless of whether it is helpful or not. I admit, I've felt the urge to click on that 'No' button when I disagree with a reviewer, but I realised that disagreeing with someone does not mean that I found the review unhelpful, and does not justify my choosing to vote 'No'. In fact, the converse could be true - I could disagree and therefore find a review helpful as it motivates me to think why I enjoyed a book, found a book interesting, or why I should go ahead and write a review of my own.

The problem, really, is one of incentives. Reviewers confront a problem: do they wish to tell their true opinion about a book and provide a service, or do they want to be highly ranked in the Amazon system, or can they do both? With the current system, you probably end up with people in the top ranks of reviewers being those who give books 5 stars regularly, with fans of the book always giving them many 'helpful' votes when they just liked the book. Which means, if you're competitive and want to rank well, there's an incentive to write five star reviews more regularly than critical or harsh reviews in case fans of a book disagree with you and destroy your ranking. Now this doesn't matter much to me, I'm small-fry in the world of Amazon reviewer rankings, but, as an economist, I can't help but think that people will turn the system into a competition over positional goods and that the incentives are inadequately geared towards truthful, systematic reviews of books. Honest and critical reviewers must either be non-competitive or have altruistic preferences, or be uninterested in, or indifferent to, their reviewer rank.

I have assumed that the two outcomes - being highly ranked and being truthfully critical - are mutually exclusive because of the potential for a critical reviewer to be swamped by the nays of fans. In fact my conclusions don't require this assumption, they probably need something weaker like an interaction of spitefulness and criticalness, which is likely to occur. I don't know if fans actually act spitefully, but the evidence in behavioral economics suggests they might. What could Amazon do to counteract this outcome? Currently, I don't know, but I think that the behavior stems from an asymmetry in the costs to reviewing and the costs to commenting. I would suspect that far more customers vote on whether they find a review helpful than write actual reviews. Therefore, maybe Amazon, to reinforce truth-telling, should only allow you to vote on a review if you yourself have written a review, say you get 5 votes per review you write. We'd then get a second order problem of review quality as everyone who wanted to vote might write poor reviews... Gosh what a nightmare.

One countervailing factor could be reputation. If a 'real name' reviewer has a reputation to uphold outside of the world of Amazon.com, then they might have the incentive to tell the truth about their opinion of a book because, if they don't, their reputation would be undermined and they could suffer pecuniary losses outside of Amazon. This is why I favour the 'real name' reviewers option, and why I would distrust anonymous reviewers.

But, here's the kicker, it isn't in Amazon's interest to get people to tell the truth about books, CDs, or DVDs. They are in the book-selling business, not the book-reviewing business. Thus, assuming there's a positive correlation between the number of stars a book receives in review and the likelihood that any person will buy it, the more people who write 5 star reviews of books, the more books Amazon will sell. Amazon's ranking system therefore fits its intentions perfectly - get people to compete over being top-ranked, which requires those people to write more 4 or 5 star reviews than fewer star reviews, which then compels people to buy more books.

I conclude then that trusting Amazon reviews, on aggregate, is probably not a good thing to do. I know that there are certain reviewers who write interesting and high quality reviews, for example Herbert Gintis's review page on Amazon.com is great, but I'm not sure if there are many Herbs out there, and therefore I'm not sure how many Amazon reviews I can trust. Amazon is in the book-selling business after all.