In Dominic Cummings and Boris Johnson’s new-look Whitehall, spin doctors, civil servants and political strategies are out and algorithms are in.
Nothing has exposed how badly that ideology is faring than the government’s disastrous exams algorithm, which tried to objectively assign grades to every student in the country. It had one crucial flaw: it was designed to ignore the damage to the life chances of individual students in favour of optimising the nation’s grade averages. After public outcry, it was scrapped. But another “mutant algorithm” haunts the halls of power: a proposed new formula to decide where to build 300,000 new homes and solve England's housing crisis.
Created by the Ministry of Housing, the algorithm tries to perfectly allocate an annual house building target to each region by factoring in demand, population growth and local affordability.
But according toanalysis from planning consultants Litchfields it is already producing some pretty odd results. Rural Tonbridge in Kent, for example, is being told to build 1,440 homes annually – enough to grow its population by just under a tenth in size every year. It’s one of a number of leafy shires around London that have been told to build thousands more homes and eat into the Green Belt. London has been told to treble the number of houses it builds despite the scarcity of building space, while the north east, north west and Yorkshire and the Humber, where there is also high demand, have all been told to reduce their output.
Beyond unhappy councillors in the Tory heartlands, if this new algorithm backfired, it could have a big impact on the economy and regional inequality and worsen an already critical shortage of affordable homes.
So where did this all begin? Like so many unpopular policies, it was created with the best of intentions. For decades, house building in England has been a bureaucratic nightmare. Council planning is infamous for being plagued by angry residents opposed to any and all developments – nicknamed NIMBYs (Not In My Back Yard). Faced with opposition from these residents, local authorities frequently struggle to even come up with a number for the new homes they need to build each year. This algorithm gives them that concrete figure.
“There’s a lot of consensus in the industry that that’s a good idea – that you need to at least get to a number,” says Stephen Gleave, who sits on the planning group for the Royal Institute of British Architects. “Local authorities were falling way behind producing their plans because there was so much argument over the basic number of houses they have to plan for.”
But in practice, there are some major consequences to the new formula, not least for England’s regions. If the algorithm has its way, London and the south east will have hugely inflated new targets for housing, while cities further north will have their targets slashed to levels lower than their current output. Even the House Building Federation, a trade group that supports much of the policy, says they “recommended changes” to the government to make sure the algorithm actually delivered “homes in the north”.
These regional inequalities arise from the constraints of the algorithm itself and its heavy weighting toward demand. So if, for example, London generally needs three times as many homes as Liverpool, then Liverpool’s share of the new housing target would roughly be a third of the capital’s, irrelevant of whether the figure matches either city’s real needs.
The government chose 300,000 homes a year as an achievable target for the country, but the real demand could be closer to 400,000 or 500,000 according to Jonathan Webb, a housing researcher for the Institute of Public Policy Research. “The problem with this algorithm is it’s so constrained by what the government perceives the housing need to be,” says Webb. “But that target is so short of what is actually needed that when you try to produce enough supply in London you end up reducing the number of homes you build in other areas. And the problem with that is those other areas still need homes.”
Take Manchester, which is currently in the midst of a housing boom. The homes being built are expensive – between 2016 and 2018, not a single one of the city’s 14,667 newly built homes was classed as “affordable”. The problem is so acute it’s become the basis of a new BBC TV show. But the new algorithm isn’t just telling the city not to produce more affordable housing, but to actually produce 31 per cent fewer new homes each year.
These reductions can have ruinous consequences for the local housing and development sectors. The construction industry in growing northern cities contributes hundreds of millions to the local economy – in Manchester alone it is estimated to be worth over £4 billion and is the city’s fastest growing sector.
“We know that house building is a massive benefit to the economy and there’s a lot of money tied up in it. It involves a lot of different industries,” Webb says. He claims that the whole sector in certain regions could be at risk if the government lowers the priority of development in those areas. While it’s still uncertain whether developers will be stopped from building extra homes in these northern cities, it seems likely that government support for the industry will be given a significantly lower priority than it is now. Everything from tax incentives and funding to fast-tracked permissions and joint projects could be at risk and the effect that would have on the local economy is devastating.
And just as there are losers under this algorithm, there are winners too. With a huge drive for housing in certain regions, the demand for new land to develop will skyrocket. As a result, the value of the finite amount of land in those areas will also shoot up. Land is increasingly something that is owned by a select few (half of the country is currently owned by less than one per cent of the population) and many of those landowners are set to do very well from any concentrated surge in house building. In places like Camden in London, which has been told to increase the number of homes it builds each year from under 1,000 to 5,604, there is a very small amount of land available. Even if the government’s newly proposed planning reforms manage to release more land for development, it’s unlikely they could ever free up enough to account for a surge in demand that high. This kind of land inflation has other knock on effects too.
“If the cost of land is still extortionately high, then who is going to be able to afford to build those homes? Chances are it’s going to be the sort of people who know they’re going to be able to make a profit on it,” says Webb. In essence, the more expensive the land, the more likely that housing built on it will be luxury and high-end – the types of buildings that give developers bigger margins, defeating the point of the algorithm entirely.
Many of these fundamental problems are emerging even though the algorithm isn’t even active yet. It’s still in the early white paper stage, meaning it’s far from even a parliamentary vote let alone being officially implemented. Experts and the housing industry are confused about the exact ramifications of the new housing targets. Many say they aren’t sure what role the government would play in supporting or undercutting local development in affected areas.
The Ministry of Housing says it is happy to “update or refine” the algorithm to help handle potential complaints and politicians and industry figures are already trying to fiddle with its calculations.
Earlier this month, London mayor Sadiq Khan offered his tentative support to the algorithm on the condition that different mathematical weighting is applied to affordability and housing stock calculations – in essence changing the formula to be more suited to the economic situation in London. Given this government’s history of U-turns, it doesn’t seem out of the question that the algorithm could be significantly altered before being implemented.
“In these formulas you end up manipulating the number to something which comes down to a whole variety of other factors like politics, acceptance, local community belief or ambition,” says Gleave. “This is just juggling to fix the algorithm on a number that is intuitively and pragmatically better and undermines this idea that you can just come up with an answer from a formula and ignore the local situation.”
The housing algorithm and the exams algorithm are hardly the first of their kind. Earlier this year, police forces in Britain pressed ahead with widespread trials of facial recognition software despite the fact the technology has been found to have wrongly identified people as much as 98 per cent of the time.
Meanwhile crime prediction software, which attempts to work out if an individual will commit an offence in advance, had been adopted by 14 police forces as of last year. This is despite the fact that the software is allegedly unfairly biased against lower-income and BAME communities.
The recurring theme here is an assumption that by just using an algorithm you can find a completely objective solution to any issue. That all these algorithms have struggled as they come into contact with the real world suggests otherwise. “This is essentially the problem with all algorithms,” Webb says. “They just reproduce human biases.”
This article was originally published by WIRED UK
