Consumer Policy Research Centre Chief Executive Officer Lauren Solomon reflects on the need to broaden our understanding of consumer data issues in Australia, beyond establishing a Consumer Data Right and the Review into Open Banking.
“If you are using a product or service and not paying for it (or sometimes even when you are), then you are the product.” — Productivity Commission, Data Availability & Use Inquiry, Final Report (2017).
Data as a tool to drive competition
Opening up consumer data can be an important step towards enabling greater competition, facilitating comparison and improving product choice. Establishment of a Consumer Data Right in Australia is the first step towards providing consumers greater ownership and control over their data. The Federal Government kicked off this process last Friday, releasing the Review into Open Banking Final Report for consultation. In response, UNSW legal researcher Dr Katharine Kemp has provided some particularly insightful commentary raising potential data privacy, consent and discrimination issues.
You can tell an awful lot about someone from their transaction and consumption data. Combine this information with, for example, browsing history, social network, or location data and, suddenly, you can know more about someone than close friends and family do. Companies like Acxiom have now collected 1500 data points on 96% of the American population.
And herein lies one of the big emerging issues for the new world of data amalgamation.
Firms are making significant profits from amalgamating large amounts of different data. Buried deep within terms and conditions are clauses that give consent not only for companies to share the data about our activities and behaviour, but also to acquire more data about us held with other companies. When this data is amalgamated, it creates the potential for greater information asymmetry – with the scales increasingly tipping in the seller’s favour.
Information asymmetry & pricing
One fundamental criterion of perfect competition is perfect information: that both buyer and seller have perfect or complete information about the transaction. George Akerlof (1970) famously highlighted the role of imperfect information, or information asymmetry in his analysis of the used car market. When sellers have more information than buyers about the transaction, Akerlof showed this can lead to “adverse selection”, with buyers often ending up with a lemon of a used car.
Consumer data, when amalgamated, can absolutely increase the knowledge of the seller, which would suggest an increased likelihood of adverse selection due to the knowledge and power imbalances inherent in the trade.
When a company knows a great deal about our profile, they can also better engage in what’s known as price discrimination: charging different consumers different prices for the same product with the same cost, or, more accurately, when two products are sold at prices that are in different ratios to their marginal costs (Stigler, 1987).
In the new age of big data, companies can now “screen” customers by relying on a profile that has been built of that consumer using proxies such as their credit score, payment history, or browsing history. Big data and the potential for greater price discrimination has been raised in relation to health insurance with warnings from the Actuaries Institute, airline ticketing, and even Amazon DVD sales. Just last week, we read reports of genetic testing being mooted to screen consumers for health insurance products.
Price discrimination, however, is not the focus of this article. What’s perhaps more important is how the algorithms that determine our profiles and scores are being developed. Do consumers get access to their scores and profiles? What are they built from? How do we correct any errors within them, or change our behaviour to improve our profiles and scores?
What really goes into developing consumer profiles or scores?
If our profiles are used to limit or alter the choice of products we are presented with, how can we gain access to these profiles or scores? Reforms to date in Australia have mainly focused on Australians securing access to their own data. The line has been drawn short of giving us access to the profiles which have been developed of us, or what types of data went into making that profile or score.
This is concerning for several reasons, but I’ll explore two key issues.
Firstly, every undergraduate economics student is taught that a model is only as good as the data you put into it. Or, as I recall my lecturer putting it, “garbage in, garbage out”. If a score or profile is based on incorrect or irrelevant proxies, or spurious correlation, then that’s an issue. Without transparency over what data has gone into constructing their profile, consumers are also not able either to adjust their behaviour, or challenge a decision over access to a product or service. Ultimately, and over the longer term, this has the potential to erode trust in the market itself, particularly if such practices become widespread.
Secondly, international experts such as mathematician Cathy O’Neil have been calling for caution precisely because these algorithms can reinforce existing biases and discrimination within society. A predictive model that uses historical data and proxies to make predictions about future behaviour or value of the consumer, embeds the disadvantage and inequality already inherent in society. This worldwide phenomenon is giving rise to organisations like the Algorithmic Justice League and the UK Data Justice Lab. These teams of data scientists, ethicists and privacy experts are exploring ways to deliver a more inclusive society, a focus which I believe should also be brought into the consumer policy space.
It’s not all good, or bad … but it’s not neutral
You’ll see many people writing on these issues referring to Kranzberg’s First Law – “Technology is neither good nor bad; nor is it neutral”. The reason I’m also drawing on it here is because it’s perhaps the best reminder that, in a world of fast-changing technology, doing nothing in policy terms isn’t neutral anymore – doing nothing has consequences.
Part of the challenge with this field, I believe, is that it crosses the disciplines. It pushes the boundaries of old assumptions which previously, within each field, might have been broadly accepted. Price discrimination on its own is not new. Screening on its own is not new. Coding is not new. Behavioural interventions are not new. What is new is the supercharged computing power and data amalgamation occurring across multiple sources and platforms at a level we’ve not seen before.
As James Plunkett for Citizen’s Advice recently highlighted, the trends are clear – computing power will increase; big data will become more powerful; consumer markets are increasingly going digital and online. Without appropriate adjustments to the regulatory and policy framework – such as increased price monitoring, and complementary policies to protect against exclusionary practices, especially in essential service markets – we risk increasing inequality and further eroding consumer trust.
CPRC is putting consumer data at the centre of our research agenda over the coming few years. We’re welcoming collaboration with organisations and experts across a range of disciplines to explore this new field and to ensure that consumers are best placed to secure the benefits from these new technologies and reforms.
While our markets are undergoing such major digital transformation, we simply cannot afford to be stuck in policy neutral.
Stigler, G. (1987). Theory of Price. Macmillan, New York.
Consumer Policy Research Centre is an independent consumer think tank established by the Victorian Government in December 2016. We undertake interdisciplinary and cross-sectoral research to inform government policy and business practice change. Our goal is to deliver a fair outcome for all consumers.