Transcript
Check against delivery
Good afternoon ladies and gentlemen and thank you for the opportunity to address an intellectually fascinating topic, and an increasingly important one.
This topic goes to the heart of competition policy and how the ACCC will prepare and protect markets from anti-competitive and anti-consumer influences.
We are now, clearly, living through another economic revolution – the information revolution. While perhaps a long time coming, it has now arrived.
As we all know, vastly improved data capture, processing speed and developments in machine learning are reshaping the way our economy functions before our eyes. And each movement leaves a crumb which says not only where you’ve been but also where you’re likely to go next.
The volume of data captured and stored is increasing rapidly.
The Productivity Commission’s recent report into Data Availability and Use highlights estimates that 90% of the world’s information has been generated in just the past two years.
While not all data is used or usable, a vast amount still is being used, with structured data[1] accounting for about 20% of all data.[2]
The variety of data has also increased as the growing internet of things tracks customers’ behaviour more closely than would have been thought possible, and perhaps reasonable, ten years ago.
As important, the velocity at which data can be accessed and processed is increasing, giving firms the ability to forecast how an event will unfold in real time.
This process is often referred to as data-driven innovation.[3]
Today, I will draw out three points about the revolution we’re living through.
First, there are enormous benefits and opportunities from data-driven innovation. Handled correctly, the upside dwarfs any downside, especially when your focus is on competition and consumers.
Second, and as the paper by Stucke and Exrachi demonstrates, competition issues may arise, but with the Harper changes which came into force on 6 November, I believe they can now be addressed.
Third, the Productivity Commission’s recent report on data sets Australia up very well to reap the rewards and limit any downsides.
1. Benefits and opportunities of data-driven innovation
The ACCC sees many economic advantages, realised and unrealised, in data-driven innovation.
An obvious benefit is the ability for customers to rapidly find what they need online. This reduces search costs and can increase competition between goods and services providers.
More often than not, data-driven innovation also develops efficient solutions to everyday problems.
To use transport as an example, most of us carry a phone that identifies navigation and transport options in real time. We can compare modes of transport to take us from A to B based on likely travel time and price.
When driving, apps on our phone can identify traffic issues in real time, suggesting we take a different route given conditions on a given day.
The Productivity Commission has, in last month’s excellent report Shifting the Dial, identified many unrealised opportunities for government and business in Australia to make better use of data to improve both our economy and government services.[4]
This is being done with some success, at relatively low cost, overseas.
For example, Transport London recently used Wi-Fi and data analytics to look for ways of improving their network.
In just four weeks, Transport London collected data from 5.6 million mobile devices detailing 42 million journeys, enabling them to look at how passengers travel within and between stations.
This new knowledge helped them understand how to address congestion within and around stations, and to identify commercial opportunities to promote new high traffic locations to tenants and advertisers.[5]
Data-driven innovation also allows efficient price discrimination. Uber fares on New Year’s Eve, while controversial, lead to supply being available when it would otherwise not be.
And petrol price apps are only going to improve beyond helping consumers work out when and where to buy, to allowing a wide range of other offers.
The ACCC (obviously) strongly welcomes all these innovations.
2. Potential competition issues
However, these developments clearly have many consequences for markets, and the ACCC is considering cases where algorithms are deployed as a tool to facilitate conduct which may contravene Australian competition law.
Within our organisation, we are building the expertise to analyse algorithms.
The ACCC now has a Data Analytics Unit which we use in market studies and to support the work of our investigations teams and economists.
And to stay abreast of developments, we are engaging with other competition authorities and practitioners about these issues.
Algorithms and artificial intelligence have been a recent focus of the OECD Competition Committee, and the ACCC is co-chair of the International Competition Network’s Unilateral Conduct Working Group, which has been focussing on online competition issues for some time.
Through this work, we have identified a number of areas where competition issues may arise. The most obvious is the market power of online platforms.
Last year, following an ACCC investigation into the vertical restraints imposed by online travel agents, Expedia and Booking.com agreed to remove contractual requirements restricting Australian accommodation providers from offering better room rates or different inventories to other online agents and through offline channels.
Given the dynamics of some technology markets, the ACCC was content to complete these investigations with speedy administrative resolutions.
However, it would be wrong to assume such resolutions will always satisfy the ACCC. And it is worth noting that the ACCC can revisit these resolutions if they prove ineffective. Following the High Court’s Flight Centre ruling, we announced that we are again looking into aspects of the conditions imposed by online travel agents in their contracts with accommodation providers.
Big data and mergers
It is not uncommon for the first movers who develop the networks and collect the data to be rewarded for taking on the risk and investment by being only one of a small number of providers with access to certain data.
This may not, of itself, raise competition concerns.
However, in the merger review context we will carefully consider acquisitions where both parties are involved in collecting and selling big data, or they are vertically linked in the big data supply chain. Obviously, the competition issues raised will vary on a case-by-case basis and depending on the market involved.
Like any valuable asset, the collection and possession of big data may be a decisive factor in certain transactions. For example, this may be the case where a merger is likely to have the effect of foreclosing access to unique data that is essential for competitors to compete or for new rivals to enter the market.
As another example, it could be argued that the acquisitions of a maverick firm may be more harmful in a market where big data would otherwise facilitate coordination.
While we expect that our standard analytical framework will apply to big data just as it does in other mergers, we recognise that big data may present certain challenges, including in assessing the future impact and value of that data within a rapidly changing technology environment.
Price algorithms and collusion
A key area where competition issues may arise is the subject of today’s panel: namely whether data driven innovation increases the risk of collusion.
It is of course accepted that collusion leads to poor economic outcomes.
On the other hand, mere parallel conduct is considered a rational response to market conditions, and so is permitted. Something more is needed to warrant enforcement intervention by the ACCC.
Cases brought to date globally by competition authorities relating to the use or misuse of online databases to determine prices reflect circumstances where “something more” occurred.
For example, in the United States Airline Tariff case,[6] a database that was generally available to travel agents was used by airlines to negotiate supra-competitive airfares and ensure that proposed price rises stuck.
As another example, the ACCC’s action against Informed Sources and petrol retailers focussed on the agreements to exchange price information between petrol retailers and Informed Sources.
It was alleged that those agreements, in combination, provided fuel retailers almost real time access to price changes by their competitors that would not otherwise have been available, and which the ACCC alleged substantially lessened competition between those retailers.[7]
Concern has been raised by some that the way prices are determined, and potentially collusive outcomes are achieved, is changed by machine learning algorithms.
It is argued that, in the right market conditions, pricing algorithms may be used to more effectively engage in and sustain collusion, whether ‘tacit’ or not, reducing competition but without contravening competition laws.
It is said that a profit maximising algorithm will work out the oligopolistic pricing game and, being logical and less prone to flights of fancy, stick to it.
To further complicate matters, the development of deep-learning and artificial intelligence may mean that companies will not necessarily know how, or why, a machine came to a particular conclusion.
To this end, it is argued that if similar algorithms are deployed by competing companies, an anti-competitive equilibrium may be achieved without contravening competition laws.
Contrary views have also been put.
Critics say that models predicting this outcome require a number of elements, including limited brand differentiation, to stick.
They wonder whether the increased price discrimination made possible by big data analytics will make collusion harder, rather than easier.
They say there is no evidence at this stage that the collusive result would be any worse than the current status quo in oligopolistic markets.
And they note that it is not clear how collusive algorithms will respond to repel new entry without contravening existing competition laws.
I will be interested to hear from the experts on this today. However, there is one observation I will make in relation to this discussion. In Australia, we take the view that you cannot avoid liability by saying “my robot did it”.
Algorithms and liability
In Australia we have some authority for this.
Some in the audience may recall the ACCC’s Google AdWords case, which went to the High Court in 2013. In that case, the ACCC was concerned about misleading search results produced in what were then known as Google’s “sponsored links”. These arose between 2005 and 2008 when businesses added their competitors’ names to Google’s advertising algorithm to ensure that when a consumer searched for, for example, “Harvey World Travel”, the consumer saw a sponsored link to “STA Travel” at the top of the search results.
The courts’ approach to attributing liability in the Google AdWords case may be instructive today.
The Google AdWords case is probably best remembered for the High Court holding unanimously that Google did not create or produce any of the sponsored links, and therefore was not liable for the misleading conduct.
It is sometimes forgotten that the Court did find that the sponsored links were misleading.
It is also sometimes forgotten that the businesses that used Google’s algorithm to create these misleading links were found to have contravened Australia’s consumer laws.
Finally, the High Court decision noted that there was no allegation that Google was “knowingly concerned” and, by so doing, left open the possibility that, had it been pleaded, Google may have been found liable on that basis. Subsequently Google made sure the practice no longer occurred worldwide.
But what happens if an artificially intelligent robot engages in sustained collusion with another robot, either through the “predictable agent” or “autonomous machine” scenarios posted by Stucke and Ezrachi. My answer is ... let’s wait and see.
However, I am confident that our laws, particularly with the addition of the new concerted practices prohibition but also the new misuse of market power provisions, can deal with either situation if they substantially lessen competition. At this stage, the ACCC has not seen any anti-competitive algorithms which require an enforcement response beyond what is now available to the ACCC under Australian law.
It is very likely, however, that big data and algorithms will expand the activities that give rise to concerted practices.
The recent Harper changes are important here.
Let me give you an example … Assume a machine-learning algorithm is deployed by a firm with substantial market power to determine profit maximising downstream prices and that machine determines the best course of action is to impose a margin squeeze on the firm’s downstream competitors.
Would this contravene the old misuse of market power provision?
It may be difficult to establish that a firm with substantial market power had a proscribed anti-competitive purpose when deploying that algorithm. By focusing on the effect or likely effect of conduct, however, the new misuse of market power provision is fit-for-purpose to prohibit this conduct.
Similarly, the new concerted practices prohibition should help shift the focus away from a requirement to establish a “meeting of the minds” to consider whether these has been cooperation between competing businesses that substantially lessens competition.
If robots are colluding, this provision will help us stop this conduct.
I expect that, no matter how anti-competitive conduct occurs, there will now be a legal hook allowing the ACCC to take appropriate enforcement action to address the resulting competitive detriment.
If I am not correct and the law is ultimately found wanting, I also have little doubt it will be amended, if only because the implications in the new data-rich world are so far reaching.
Finally, as with any area of commerce, there remains an important role for consumer protection. While often old school issues, vices can arise in new ways with new technology. Just last week, the ACCC received judgment in its action against Meriton Property Services Pty Ltd, with the court finding that Meriton engaged in misleading or deceptive conduct and conduct liable to mislead the public by taking steps to prevent guests it suspected would give an unfavourable review from receiving TripAdvisor’s ‘Review Express’ prompt email.
Other current ACCC cases and investigations involve the way businesses treat consumers access to data about them or their products, or the way they use online access to impact the consumer’s enjoyment of products bought.
We will also be concerned by misrepresentations that mislead consumers into giving away their data.
3. Australia could soon be set up well to limit the downsides of ubiquitous data
One issue with digital devices is that, as The Economist recently observed:
“Buyers should be aware that some of their most basic property rights are under threat … they should fight for the right to tinker with their own property, modify it if they wish and control who uses the data that it hoovers up.”[8]
The Economist cites Tesla drivers who were prohibited by Elon Musk from working for Uber, John Deere’s requirement that only authorised software be allowed, and various goods being increasingly hard to fix.
The ACCC is already dealing with some of these matters. For example, our case against Apple where we allege that Apple denied fixes to a fault triggered by its software when it found non-authorised repairs; and our motor vehicles market study is recommending that manufacturers provide independent repairers access to the same technical data they provide their dealers.
There is now, I believe, further reason in Australia why individuals should see data as an opportunity.
The Productivity Commission is ahead of the game by calling for consumer empowerment in the use of data. The Productivity Commission has called for us to move from a system based on risk aversion and avoidance, to one based on transparency and confidence in data processes.[9]
That is, to treat data as an asset rather than something personal that must be protected.
The Productivity Commission has described this as a new competition policy, driven by a right to use your data. It is essential that access to data, while underpinned by confidence in privacy, is considered through a competition and consumer lens to see the paradigm shift envisaged by the PC.
The right would be inalienable; not able to be contracted out or sold. Importantly, the Productivity Commission review sees consumer data operating as a joint asset between the consumer and the entity holding the data. Both parties would have the opportunity to harness the dataset for their own purposes, and consumers may share their consumer data with third parties.
Harnessed effectively, this will enable consumers to better compare products and services, facilitate switching and may help address some barriers to entry.
This is not to understate the challenges ahead.
The Productivity Commission states unequivocally that “marginal changes to existing structures and legislation will not suffice”[10] and consumers would need to embrace any such right. The ACCC considers that while industry and experts might be best placed to start the journey, it will need strong legislative backing and a regulator prepared to put competition and consumer issues to the forefront.
There are numerous reasons why such a policy shift is timely.
Finally, ACCC market studies, such as our communications sector market study, are powerful tools. By focussing on a specific sector, we can sort through the white noise to determine if there are competition issues that should be addressed by way of an enforcement or policy response.
As you will know, the ACCC may soon be conducting a market study into the impact of the new digital environment on media. We will go into this with an open mind but we do believe that the market position of these platforms warrants a detailed assessment to determine the influence and impact of these platforms on content creators, advertisers and consumers.
While the inquiry is mainly to make recommendations to government, we will also be alive to any potential breaches of our newly invigorated misuse of market power provisions.
Thank you for your time today.
[1]‘Structured data’ refers to information with a high degree of organization, such that inclusion in a relational database is seamless and readily searchable by simple, straightforward search engine algorithms or other search operations; whereas unstructured data is essentially the opposite.
[2] As cited in Productivity Commission Inquiry Report, No. 82, 31 March 2017, page 57
[3] For example see OECD, Data-Driven Innovation – Big Data for Growth and Wellbeing (2015)
[4] Op Cit. 1.
[5] Luke Upton, https://www.smartrailworld.com/wi-fi-use-big-data-analytics-better-passenger-journeys-in-london 11 September 2017
[6] United States v Airline Tariff Publishing Co., 836 F. Supp. 9, 12 (D.D.C. 1993).
[7] The ACCC resolved these proceedings in December 2015: https://www.accc.gov.au/media-release/petrol-price-information-sharing-proceedings-resolved; https://www.accc.gov.au/media-release/accc-and-coles-express-resolve-petrol-price-information-sharing-proceedings.
[8] The Economist, ‘How digital devices challenge the nature of ownership’, 30 September 2017. https://www.economist.com/news/leaders/21729745-and-threaten-property-rights-digital-age-how-digital-devices-challenge-nature.
[9] Productivity Commission, Data Availability and Use, page 2
[10] Ibid.