Since You Asked

Everyone’s a critic now – or are they?

‘The ASA has banned TripAdvisor UK from claiming or implying that “all the reviews that appeared on the website were from real travellers, or were honest, real or trusted” on its UK site’

Financial Times, February 1

The ASA?

The Advertising Standards Authority. They don’t like the fact that TripAdvisor, a company that operates websites providing travel advice, has been claiming that reviews on its UK site are written by real people.

Who else might they be written by?

Well, they could be written by automated software. Or by real people who are nevertheless not genuine customers – for example, the owner of the restaurant in question, or the owner of the restaurant across the street from the restaurant in question. Or they could be written by customers who have been offered bribes. It may well be that such abuses are rare – TripAdvisor is at pains to assert that they have various safeguards. Still, it cannot prove that dishonest or fake reviews are impossible, hence the ASA ruling.

But it’s not really news that people do dubious things online, is it?

No. And online reviews aren’t new either. But they seem to have been more of a sore point on TripAdvisor than reviews on Amazon or ratings on Ebay. Partly it’s about the stakes: from the customer’s point of view, buying a substandard romantic novel is trivial, whereas a ruined weekend tryst in Paris is a minor tragedy. It’s also about the possibility of malicious reviews. I could post grim reviews of Freakonomics and Outliers in the hope of boosting my own book sales, I suppose, but even if my reviews were taken seriously it is not clear I would benefit. On the other hand, if I ran a gastropub in a picturesque village with a single rival, it’s fairly clear that I might benefit if I could persuade potential visitors that my rival left a trail of disgusted customers in its wake. So both the customers who read reviews on TripAdvisor and the businesses being reviewed have a lot to lose if the review system is corrupt.

Full of sound and fury, signifying nothing?

Signifying $4.4bn, if TripAdvisor’s market capitalisation is anything to go by. And people do take online reviews about as seriously as they deserve to be taken.

How seriously is that?

The limited evidence we have suggests that people are drawing quite sensible inferences from the reviews they read. Ebay auctions have been studied by the economist David Reiley and a number of his colleagues. Reiley finds that positive ratings for a seller push up the prices she receives, but that the effect is small and not statistically significant. On the other hand, negative ratings have a much larger and statistically significant effect in depressing prices. This makes sense: sellers might go to some lengths to round up friends (or “sock-puppets” – online aliases that they control) and ask them to post positive ratings. But who bothers to post negative Ebay ratings for the sheer joy of it? Negative ratings are also rarer and so might be regarded as more informative.

Seller ratings aren’t the same as reviews.

No, but we also have evidence on Amazon reviews, thanks to work by Judith Chevalier and Dina Mayzlin. They compared books for sale on the Amazon and Barnes & Noble websites, observing both the reviews and the relative popularity of any given title on each site. This is a nice statistical tool. Obviously one would expect good books to earn good reviews and lots of sales, but when a book has particularly notable reviews on one site, Chevalier and Mayzlin were able to use that fact to track the effect of the reviews on the book’s sales.

And?

It’s a similar story to that which David Reiley found on Ebay: reviews can affect sales, for good or ill, but negative reviews have a much greater impact.

Wouldn’t it be better if reviewers themselves were properly identified?

This does happen – for instance Amazon can use credit card details to verify the identity of a reviewer, should that reviewer want to be identified. Social networks such as Facebook and Google Plus prevent anonymity and have tried to prevent pseudonyms too. The advantages of this are obvious as far as the likes of TripAdvisor are concerned – so too are the advantages to, for instance, the Chinese police. An alternative is to allow users of a website to review the reviewers. But who reviews the reviewers of the reviewers? Apparently it’s reviewers all the way down.

Also published at ft.com.

4th of February, 2012Since You Asked • Comments off