The value of the property.
The building’s square footage.
The age of the building.
The timeliness of its tax records.
Whether its utility bills are in arrears.
Taken together - with some clever maths applied - these are the predictive indicators for identifying some of the most dangerous buildings in New York City.
Having made his fortune providing data-driven analytics for the financial sector, as Mayor, Michael Bloomberg wanted to prove that the same approach could work in cities, too. His insight was to create the Mayor’s Office of Data Analytics (MODA). The team’s work on illegal conversions is indicative of the impact it can have.
New York responds to nearly 20,000 complaints of illegal apartment conversions every year. These are buildings that have been illegally subdivided by rogue landlords. The buildings are over-crowded. They are health and fire hazards. People get ill in them. Sometimes they even die in them.
It used to be that such buildings were inspected in the same order as complaints came it. In approximately 8% of visits by safety inspectors was an illegal apartment actually discovered. Using a data-driven approach using the predictive indicators above, it’s now consistently over 70%. That’s more than an eight-fold increase in efficiency that saves lives.
By combining, analysing and seeking insights from data from numerous different sources, the MODA team has applied the same technique to improve emergency response times, predict where fires are most likely to occur, and kept one of the world’s most complex urban sewer systems running.
The reason for highlighting this case study is that it is easy to be sceptical about the potential for big data to improve government. We have all become jaded having read innumerable articles with formulaic titles along the following lines:
'[X] reasons why big data can help [insert random organisation] achieve more [insert random benefit]'
Big Data has joined that long list of technical terms - along with the ‘Internet of Things’ and ‘Smart Cities’ – that are so ambiguous as to risk being essentially devoid of meaning. The lesson from New York is that with big data it’s not the technique that matters. It’s not the process, either. It’s the outcome. New York City has been successful with big data precisely because it has focused on delivering tangible outcomes that really matter.
The big question on this side of the pond is: what outcomes matter to us?
The UK context
While it may still be premature to predict the outcome of the next general election with any certainty, one thing seems sure: public services will have to survive on smaller budgets. Money is certainly not the be-all-and-end-all for public services - indeed, the past few years have enabled much-needed reform. Yet we have to acknowledge that we’ve reached the end of the road for merely salami-slicing back on services. Even after all projected savings over the next parliament (and having already made £10 billion of savings during this parliament) the local government sector in England and Wales faces a funding shortfall of £12.4 billion by 2020. Public sector organisations face a tough choice: either they will have to stop offering some services altogether or fundamentally redesign them.
The lesson from New York is that big data can help achieve the latter. How? By following the age-old medical cliché: prevention is better than cure. Think about the current business model of public services here in the UK. That model is to resolve problems after they have already escalated in cost and severity. We visit our doctor only once we are already ill. Councils respond to fly tipping when there’s already rubbish on the ground. We intervene to support troubled families when they have already ticked boxes in the categories of unemployment, criminal behaviour and school truancy.
We have to shift to a business model that sounds rather Blairite: We need to be tough on failure and tough on the causes of failure. That is precisely what big data offers to do: to predict where things will happen.
Imagine if we could re-gear public services to predict and prioritise problems so they could be treated when they have caused minimal harm and when they are at their least expensive to resolve. Imagine if our smart wrist-bands could notify our doctor of our high blood pressure before we got ill. Imagine if councils could use big data to realise that the reason there was fly tipping in a certain place was because the land was owned by an absentee landlord. Imagine if we could support the UK’s most vulnerable families before unemployment, crime and truancy blight their lives.
These things can be enabled by data. The public sector has mountains of it. So where should it start?
Lessons from the private sector?
The common wisdom is that 'the public sector could be transformed, if only it could learn from the private sector’s approach to data and technology innovation.' If that’s true, then we cannot ignore the fact that the greatest innovations in the private sector have arguably come through applying big data analytics to personal data. Think of the Tesco ClubCard vouchers that pre-empt your needs in a way is that often delightful and occasionally alarming. Think of Amazon and its uncanny ability to nudge you into buying that additional item you never thought you needed. Think of Google or Facebook who seem to have guessed exactly where you want to go on holiday this year.
As consumers, we accept all of this as wonderful convenience. And yet we (or is it perhaps the media?) react in horror when the public sector seeks to apply the self-same principles. Care.data taught us that. National ID cards taught us that. The proposed release of HMRC records provoked the same reaction.
What explains these double standards? Perhaps three reasons.
First, when we use a private company, we get an immediate, direct, personal, tangible reward as a result. That was not the case in the any of the public sector examples. Second, we can choose to use or not use a service offered by a company. The government, by contrast, does not choose its customers. Third, our different attitudes point to the fact that technology has simply raced beyond society’s ability develop the social norms to respond.
The example of the Samaritans Radar app is illustrative. In November 2014, Samaritans launched an app that enabled users to be alerted when someone they followed expressed a sentiment suggestive of emotional distress or suicidal feelings. It was intended to make sure that in the course of a hectic day you wouldn’t miss a cry for help from a friend. And yet the privacy lobby forced its withdrawal claiming it was a gross violation of people’s privacy. Yet all tweets are public - how can we claim privacy on a public platform? Whatever the rights and wrongs of this specific example, it emphasises the point: technology has raced beyond society’s ability to know how to respond.
So what do we do?
Do we forever bar the public sector from enjoying the same tools that have transformed private enterprise? No. But equally, it looks likely that government will come unstuck time and time again until two things happen:
1) It has a Code of Responsible Data Analytics. An Office of Data Responsibility should be established as an extension to the work of the Information Commissioner’s Office to write it and adjudicate it.
2) We give citizens a meaningful way to have control of their data. For that, the government should commit to public sector-wide compatibility with personal data stores.
So if the public sector really wants to make meaningful progress with big data right now, it should start - as New York has done - by using it for the innumerable things that don’t involve personal data. If they can prove that big data can genuinely transform public services it will be easier to make the case to the public that it can also work with more sensitive matters.
In short, we should look to New York and how it has predicted fires, saved lives and made the city run more efficiently in the face of budget cuts. The UK public sector should learn from it. The private sector needs to help them do it.
This article was first published on the Policy Exchange website.
The views expressed in the Opinion section of StatsLife are solely those of the original authors and other contributors. These views and opinions do not necessarily represent those of The Royal Statistical Society.