Economic Theory

Internet search engines are not utilities

A new study about supposed anti-competitive practices in the online search market, co-authored by net neutrality guru Tim Wu, has made headlines in recent days. The study contains a number of claims which are, from an economic perspective, unconvincing. In the context of a high-profile probe against Google by the European Commission, they could well motivate unjustified intervention by competition authorities, fundamentally threatening continued innovation on the net.

The authors take a stab at Google’s “universal search” policy, which tends to rank results from its own specialist (a.k.a. vertical) search engines ahead of those of vertical competitors such as Yelp. Wu et al. claim that this practice “is a form of self-dealing that excludes competitors.” Yet, using the terms ‘self-dealing’ and ‘exclusion’ in this context is deceptive. Both terms are used in competition policy – particularly with regard to utilities – to describe the practice of disabling consumer access to alternative providers. The classic example is an electric utility which also owns the grid, and prevents competing electricity providers from using the grid to offer their services to customers. Such a practice of excluding competing providers from essential access to a resource constitutes self-dealing, and it can be harmful to consumers because it enables the resulting monopoly to raise prices above what would prevail under competitive conditions.

But search engines are not like utility networks. The latter are natural monopolies – you are only connected to one electricity grid and one sewer system, and it would not make much sense to install parallel electricity grids and parallel sewer systems. But you can use as many parallel search engines as you can find, and it only takes a mouse click.

On a more specific note: What “universal search” does is offer a – clearly separate – set of results drawn from Google’s own services, on top of and more visible than the regular ‘organic’ answers. (The latter usually include its vertical competitors because their results are highly relevant.) This may be preferential treatment, but it certainly does not constitute exclusion. Indeed, it is a common practice in all lines of business, e.g. when Sainsbury’s places its own in-house brand toilet paper on a more visible or convenient shelf than other toilet paper brands. Also, if you walk into a supermarket asking for cough syrup, and the supermarket happens to sell it, you would expect them to indicate this to you before they point you to the pharmacy next door. And legal experts have confirmed that no EU business is required by competition rules to give identical treatment to its own products and those of competitors.

The authors go on to claim that Google’s results-ranking policy harms consumers. As evidence, they use an experiment looking at click-through rates – i.e. the percentage of users clicking on a given set of links – for the first seven links provided, according to two methodologies: the first one is Google’s current universal search, while the second one replicates Google’s organic algorithm, roughly the way the results would be ranked if Google’s own links were not posted at the top. They find that, with the first methodology, which is likely to include more Google links, 32 per cent of users clicked on one of the seven. The figure for the second method was 47 per cent. They claim that this “50 [per cent] increase in [click-through rates] is immense in the modern web industry,” and they say it is indicative of consumer harm because it shows a sizable percentage of users could not find what they were looking for when universal search was applied, compared to organic search.

But it is pointless to look at the first seven links in isolation. At no point are we told how many users found a satisfactory reply in the links below the first seven, or indeed how likely those lower results were to come from competing vertical search engines. It cannot be claimed that users are somehow prevented from access to any result other than those seven, since they can readily scroll down and click on them. In other words, restricting the analysis to such a low portion of available results defeats the entire purpose of consumer-welfare analysis, because it is not a realistic depiction of the actual process of search. The allegedly definitive figures cited in the study – 32 per cent vs. 47 per cent – are, in fact, meaningless.

Much else in the study is questionable. For instance, the queries given to subjects in the experiment (“paediatrician nyc,” “coffee louisville ky”) are typical searches on a vertical service like Yelp, which specialises in providing consumers with information about local businesses. But they are not necessarily representative searches on Google, which users go to for much more general questions. This means that, while Yelp will almost exclusively give users the names and addresses of highly rated physicians, Google is more likely to provide a diverse set of results, including, for instance, a news article of a German paediatrician dating a New York celebrity. Anecdotes aside, to the extent that queries do not reflect what users actually search for on Google, the experiment will be biased.

But perhaps the most concerning thing about the Wu paper is the extent to which it ignores a fundamental feature of online search – namely, the subjective nature of search results. Indeed, throughout the study we read about results that are “right” or “wrong” for users. This fails to acknowledge the fact that Google’s search algorithm – in fact, any algorithm, from a horizontal or a vertical search engine – is based on subjective judgements by its designers, from the preferences of its user base to the quality of the available information, to how to filter it for a better customer experience, and many other criteria. This explains why algorithms are different for every search engine, which enables Google, Bing, Yahoo and others to compete, and over time has led to more accurate, more personalised, and better results for users. Not all users are the same – they have vastly different tastes and preferences, and that is why some of them use Yahoo over Google and vice versa.

And, because online search is, by definition, based on subjective criteria, search engines must be allowed to rank results however they see fit, if innovation in search is to thrive and consumer welfare to be enhanced over time. Google has now decided to use universal rather than organic search, presumably to give more visibility to its own specialist services. They may or may not succeed in outcompeting Yelp, TripAdvisor, and others, but the extent to which they do will be determined by users. And if users do not like the new policy, they can – and do – switch to alternatives, either other horizontal search tools, or vertical tools directly, or – increasingly – mobile apps. The online search market is increasingly seamless and competitive, with more users entering it every day, and more players emerging to cater to their needs. This is the point that competition authorities must keep in mind as they look at Google in Europe, and not wildly misleading claims about an exclusion that clearly is not happening.

Diego Zuluaga is the IEA’s International Research Fellow.

Policy Analyst at the Cato Institute's Center for Monetary and Financial Alternatives

Diego was educated at McGill University and Keble College, Oxford, from which he holds degrees in economics and finance. His policy interests are mainly in consumer finance and banking, capital markets regulation, and multi-sided markets. However, he has written on a range of economic issues including the taxation of capital income, the regulation of online platforms and the reform of electricity markets after Brexit. Diego’s articles have featured in UK and foreign outlets such as Newsweek, City AM, CapX and L’Opinion. He is also a frequent speaker on broadcast media and at public events, as well as a lecturer at the University of Buckingham.