Paste your Google Webmaster Tools verification code here

Want to tackle the scourge of fake reviews? The market can help with that

Want to tackle the scourge of fake reviews? The market can help with that
The internet has led to a massive increase in the amount of information available. Often, this is a good thing. For example, shopping around to find the cheapest price for something has become far easier. But it can have its downsides. A report last week from the consumer magazine Which highlighted one such disadvantage. An investigation claimed that the review system on parts of Amazon was being undermined by fake five-star reviews. The magazine analysed the listings of hundreds of popular tech products in 14 online categories, such as headphones and smartwatches. Researchers sorted the headphone reviews, for example, by the average scores of the brands. The first page of results – those with the highest scores – consisted almost entirely of little-known brands, with nearly 90 per cent of the reviews from unverified buyers. In other words, there was no evidence that the reviewer had ever bought the item in the first place. Companies like Amazon are well aware of these potential problems. They take steps to try to guard against them. A flurry of very good posts for a less well-known brand is one of the classic footprints which enable fake reviews to be identified. But Which suggested that the volume and variety of fake reviews was so large that the defences are currently being overwhelmed. A similar problem arose almost from the very start of email, when spam first appeared. Ever since then, a complicated evolutionary game has been played between the spammers and the spam filters. It is a game because spam wins if it gets through, and the filters win if it does not. It is evolutionary because both sides are constantly adjusting their strategies. The filters seem gradually to be getting the better of it, though I am currently being plagued by emails from China offering to sell me plastic moulds. The fake review – and more generally the fake news – problem has not been an issue for quite as long, but concern over it is growing. The instinct of many people is to reach for the law, and in particular to regulate. Set up a body, staff it with bureaucrats who of course have the public interest at heart, and the problem will be solved, goes the logic. The European Commission is a strong proponent of this approach. But there are already some good illustrations of the private sector reducing what economists describe as “reputation systems failures”. For example, a 2017 paper by Andrey Fradkin and colleagues at the MIT School of Management analysed experiments by Airbnb. A particularly successful one appears to be that of the simultaneous review: both the buyer and seller post their reviews, and only then are they allowed to see what was written about them. Not all consumers give feedback. Many who have a bad experience do not bother to rate the seller or product – they just stop buying from the platform. Platform providers therefore have a strong incentive to verify posts and encourage real reviews, perhaps using monetary payments to reduce selection bias. Just as we didn’t need to regulate against spam, given time, markets will find solutions to what is currently a pressing problem.
As published in City AM Wednesday 24th April 2019
Image: Online shopping by Maxpixel is licensed under CC0 1.0
Read More

Emojis are a better metric for wellbeing than traditional data methods

Emojis are a better metric for wellbeing than traditional data methods
HMRC’s programme to make tax digital continues to roll out.
Anyone with a small business will know about the imminent deadline of 1 April, when VAT returns become digital. Quick to seize an opportunity, several companies have developed software to ease the task. The digitisation of tax raises the wider issue of whether technology will help the Office for National Statistics (ONS) put together faster and more reliable measures of the state of the economy. The VAT returns potentially give the ONS real-time information. The current methods of constructing the national accounts – the picture of how the economy is doing – remain rooted in the twentieth century. Ron Jarmin, the deputy director of the US Census Bureau, writes in the latest issue of the top ranked Journal of Economic Perspectives that: “current measurement programs are not keeping pace with the changing economy, and current methods for collecting and disseminating statistical information are not sustainable”. For example, national accounting bodies such as the ONS and the Bureau of Economic Affairs in America still rely heavily on sample surveys for their information. Jarmin points out that surveys are encountering increasing problems. Response rates by both households and companies have declined substantially, increasing costs and threatening quality. The intellectual conservatism of outfits such as the ONS is illustrated by measurements of well-being, or happiness. Hailed as an innovation when David Cameron instructed the ONS to produce this in 2014, it is based purely on old-fashioned survey questionnaires. Economists in general are traditionally sceptical of survey-based approaches. The respondents, in the jargon of economic theory, simply state their preferences when answering a series of questions. Economists place much greater weight on preferences which are revealed by the actions which people take. In the 1980s in Britain, survey after survey showed a stated preference for higher taxes and more public spending. Yet in their actions at the ballot box, people kept electing Margaret Thatcher. They revealed a preference for the exact opposite. The online world is replete with revealed emotions. Indeed, the entirely new language of emojis has evolved to allow people to do this. Modern machine learning techniques can readily translate the text of tweets and blogs into a scientifically-based measure of wellbeing. And they can do so much faster and more reliably than the survey methods used by the ONS. Jarmin urges governmental statistical agencies to rely much more on digital information in general. He argues that material “generated from transactions, online interactions, sensors, the internet of things, and many other sources can be used to capture various aspects of economic activity”. He notes the massive increase recently in the number of economists working for tech companies in the US. Here, innovative methods of data collection and analysis are the norm. Statistical agencies such as the ONS need to show the same energy and move much more rapidly into the twenty-first century.
As published in City AM Wednesday 20th March 2019
Image: Berlin Wall by Jed Record via Flickr is licensed under CC BY-2.0
Read More