Is Open Data Compatible with Insurance?


A paper presented earlier this month at an Institute of Actuaries of Australia conference explored how well the principles of open data dovetail with the tenets of insurance underwriting. It’s a fascinating paper that I would urge you to read in full, particularly for the worked example. In the meantime, I’m going to summarise here what for me are its key points.

The authors are Zofia Bednarz and Kimberlee Weatherall from the University of Sidney, and Chris Dolman from Insurance Australia Group. The paper is centred around what’s happening in Australia, but it’s not that different from what’s happening in several other markets.

They orientate the paper around two questions:

  • can you replace existing underwriting questions with [open data] requests?
  • can you use [open data] to create new underwriting questions?

Across these two questions, they focus on a central tenet of open data, being that the consumer has the choice about whether to share their data or not. In particular, they explore the implications of two more stringent aspects of Australian open data regulations in relation to privacy regulations: data minimisation and consent.

Open Data Obligations

When considering the extent to which existing understanding questions can be replaced with open data requests, the conclusion is that market practices around the secondary use of data for underwriting purposes would be in conflict with the requirements of open data regulations on data minimisation.

And the stricter conditions for the consent being given by the consumer under open data regulations (compared with existing data protection legislation) mean that underwriters would struggle to stay within the open data regulations.

The authors consider the question of what they call optionality. This is about the consumer’s choice as to whether to share data or not.

“…if we assume that sharing one’s data through the [open data] mechanism could be construed as replacing the proposal form and questionnaire, or at least sections of it, a refusal to share the data would likely be viewed similarly to a refusal to answer any or some [significant part] of the questions in a proposal form.”

This has implications for the consumer’s obligations on representation under long standing insurance contract legislation. They go on…

“…All this could essentially make [open data] data transfer mandatory for getting an insurance quote, which is against the conception of the [open data] regime as being under the control of, and at the option of, the consumer only.”

Reasonably Needed

The Australian open data regulations contain a requirement under the data minimisation principles that data needs to be ‘reasonably needed’ for the intended purpose. Against the backdrop of insurers hoping to use open data to explore new underwriting questions, the authors pose this key question…

“Is broad and substantial insurer access to consumer data held by banks, telecommunications companies and energy companies ‘reasonably needed’, merely because an insurer wishes to use it to assess risk? Are there some circumstances which are acceptable and some not? When and why?”

The extent to which data is ‘needed’ will be influenced by how well the consumer understands why it is being asked for. Yet this obvious question arises…

“…when an individual shares [open data] such as their banking transaction history, do they really know what an insurer (or anyone else) is looking for and how it is being used?”

The consequences are significant…

“We expect that most consumers simply won’t spend the time to make specific, granular choices around the data they share. The transparency for insurers when data is shared at all is therefore significantly higher than in traditional questionnaire settings - but for insureds the process becomes more opaque.”

So just when open data was meant to give consumers a more active role in deciding what data to transfer and for what purpose, its core premise of choice and control is undermined.

Differential Impact

The paper moves on then to explore a worked example of a simple product in a simple insurance market, offered to people who may on the one hand be high risk or low risk, and may on the other hand have lots of data, data they can’t use or no data at all. Using simple market scenarios, the authors come to this stark conclusion…

“…these two features of [open data regulation] (optionality of data transfer, and non-universal ability to share that data) will create some significant social challenges in insurance underwriting, likely exacerbating issues of affordability, particularly for vulnerable customers. We note that these are core design principles of the [open data regulation].”

In other words, open data is likely to increase stratification in insurance markets. Note that this won’t just be down to the actions of one firm. It will be down to a whole market of individual firms competing for profitable business, in a way that has as its unintended consequence…

“…the disadvantage of those most vulnerable and less able to participate in the digital economy. Conduct regulation which applies only to individual firms will not be sufficient to prevent such outcomes.”

In Conclusion

Open data is an idea that looks great when written down on the back of an envelope. Harsh perhaps, but true when considered against the context of three things:

  • the terms under which it will be offered to consumers
  • the market practices and principles that insurance operates under
  • the obligations of insurers and consumers when entering into a contract of insurance

What this paper exposes are the tensions inherent even at the principles level, between open data and insurance practices and obligations. People are fond of referring to open data making the insurance market more open, when it will in fact do the opposite. On top of that increased opacity, it will exacerbate the problems that people who are vulnerable or less able to engage with the digital economy, currently face.

What are the choices then? Take away the principles underpinning open data? Or take away the access of insurers to open data? The authors take a middle ground, calling for industry, regulators and consumer groups to sit down and talk through these issues.

Will sitting down and talking about open insurance make its seeming incompatibility with insurance go away? It’s possible, for when one set of principles meets another set of principles, somethings got to stay and somethings have to give. That’s what gets you beyond this present stage of trying to fit a square peg into a round hole.