The algorithmic economy is the new digital economy, in case you hadn’t noticed. Companies use the outcome of sophisticated data analytics in order to enhance internal monitoring, better tailor marketing strategies, cut processing costs, deploy new or refurbished products and services, and determine sale prices. On their side, consumers have been embracing algorithm-based services such as search engines (e.g., Google and Bing), aggregators (Booking.com and Amazon Marketplace), rating services (e.g., Tripadvisor and Yelp), recommender systems (e.g., Deezer and Spotify) and bidding agents (e.g., eBay’s ) to help them navigate the web and make purchase decisions, both online and offline. Moreover, recent advances in the area of artificial intelligence (AI), combined with Gargantuan strides in computing storage and processing power, open new opportunities to employ algorithms along the economic value chain and to further automate market transactions. Parallel developments in the so called Internet of Things (IOT) are likely to intensify these trends. IOT devices and personal digital assistants fed by artificial intelligence and big data troves are deployed to make purchase and other decisions on the consumer’s behalf by directly communicating with other systems through the internet and, possibly, with decreasing amounts of human involvement. Sergey Brin, Google’s co-founder, recently envisaged a possible future in which automation will alleviate “some of the more mundane tasks”, allowing people to “find more and more creative and meaningful ways to spend their time.”
A recent event organized by Concurrences in London (available slides here) gathered an impressive quantum of human intelligence in order to, among other things, have a good look at the algorithmic economy through the lens of competition policy. By rooting still rather futuristic technologies in the more familiar framework of markets and economic power, the excellent conference speakers painted some vivid scenarios in which potential issues could play out. At the creepier end of the spectrum, discussions even hinted at questions reminiscent of the most intriguing sci-fi books and movies, such as whether AI-enabled, increasingly autonomous market tools could “address” the profit maximizing objective differently from how (even hyperintelligent) human agents would do.
The underlying script of at least two conference panels was largely provided
by the bedside book (in the sense of livre de chevet) “Virtual Competition”, recently written by Ariel Ezrachi
and Maurice Stucke, two of the distinguished conference speakers. While keeping abreast of fast-moving technological developments should never
be an afterthought, the Authors convincingly argue that there is a responsibility
on competition enforcers and other stakeholders to up their game in order to
meet the challenges posed by increasingly algorithm-based markets and protect consumers.
New digital technologies hold innumerable promises, but they can also present some perils, such as when firms selling online employ algorithmic tools that prevent consumers from enjoying truly competitive prices. To start with, competition authorities in the UK and in the US, in two separate investigations, found out that online sellers of posters (or wall décor) implemented anticompetitive arrangements restricting price competition by the use of automated repricing software. Between the two online sellers investigated by the CMA there was clear agreement to facilitate price coordination, with ample evidence that they aimed not to undercut each other’s prices. As to the chosen tools for ensuring price co-ordination, these were software packages offered by different providers. As explained in the Decision, the software employed by one of the sellers could be configured in such a way (“rule 55”) that her prices would match the price of the other seller (identified by the Amazon Merchant ID number) provided that a) there was no cheaper third party seller on Amazon UK and b) the price did not go below the seller’s unilaterally set minimum price. Where the price fixing arrangement did not apply, the software was configured to undercut competing products on Amazon UK Marketplace. The automated repricing software employed by the second seller was programmed in such a way that the usual undercutting rule (“compete rules”) did not apply to the first seller (“ignore list”). What Ariel Ezrachi and Maurice Stucke in their book call the "Messenger Scenario" was unanimously considered a "no-brainer" by the conference speakers. In fact, in the poster cases mentioned above, as well as in previous instances in which the collusion was facilitated and monitored by computers, human market actors are clearly liable for the restriction of competition.
While this is certainly true, and further scenarios such as the "Predictable Agent" and the "Digital Eye" pose a number of unequivocal challenges to competition enforcers (more on them later), how powerful online intermediaries are continuously reshaping competition dynamics in the digital economy is also worth the attention of many a conference panel. As marketplaces offer customers the possibility to easily compare prices of products without incurring significant cost, sellers’ pricing decisions become increasingly crucial. At the click of a mouse, sellers employing algorithmic tools can choose with whom to compete and with whom not to compete in the marketplace. Thus, for instance, a seller may decide that it is not worth competing with sellers shipping from distant regions, as many of her potential customers are likely to prefer a faster shipping time. Prices displayed on the marketplace go up and down, also many times in a single day, fluctuating according to highly inscrutable parameters. In many instances, these fluctuations are the likely result of strategic sellers trying to “game” their competitors pricing tactics (both algorithmic and non-algorithmic), but also trying to acquire higher ranking and product visibility on the platform (e.g., being featured in Amazon’s Buy Box, “the most valuable small button on the Internet today”, as more than 80% of sales on Amazon go through it). Moreover, the poster cases have shown that sellers could employ algorithmic pricing in order to achieve anticompetitive results and that, rather alarmingly, collusion can literally be one click away. For these and other reasons, there is the pressing need to better understand the behaviour of algorithmic sellers in specific competitive scenarios such as marketplaces, where the rules of the game are largely decided by the platform (“Truman Show”).
