Neil O’Brien: The next algorithm disaster – coming to a Conservative constituency near you. This time, it’s housing growth.

24 Aug

Neil O’Brien is MP for Harborough.

Algorithms have been in the news, not for good reasons. One lesson from the A-levels row is that principles which seem reasonable can lead to outcomes you don’t expect. Another algorithm’s coming down the tracks: the new formula for how many houses must be built in different places. There are few with higher stakes.

I wrote about the housing White Paper in my last column: it proposes not just to change the methodology for assessing housing need, but also to make a standard methodology compulsory for the first time. In other words, if we don’t like the results of the new algorithm, we’ll have blocked off the emergency exits.

The new algorithm is set out here. It’s not particularly easy to read. For example, one of many factors is set out in bullet point 30:

Adjustment Factor = [( Local affordability factor t = 0 – 4 4) x 0.25) + (Local affordability ratio t = 0 – Local affordability ratio t = 10) x 0.25] +1 Where t = 0 is current yearr and t = -10 is 10 years back.

Clear enough for you?

I thought it might be a while before we saw what the new algorithm would produce in practice. But Lichfields, the planning consultancy, has translated the algorithm into what it would mean for local authorities.

The numbers that the formula spits out can be compared to the number of homes actually being delivered over recent years, or to the numbers in the current (optional) national formula. Whichever way you look at it, it’s controversial.

I’ve long argued we should concentrate more development in inner urban areas, for various reasons I’ll come back to below.  But this algorithm doesn’t do that – at least not outside London.  In the capital, the algorithm would indeed increase numbers substantially.

But in the rest of England the formula takes the numbers down in labour-run urban areas, while taking them dramatically up in shire and suburban areas which tend to be conservative controlled.

Overall, the algorithm proposes a south-centric model of growth for Britain (with some growth in the midlands).

If we compare the algorithm to recent delivery, the South East has been delivering just over 39,000 homes a year, and will be expected to increase that to just over 61,000, a 57 per cent increase. The East of England would see a 43 per cent increase, the East Midlands a 33 per cent increase, the West Midlands a 25 per cent increase and the South West a 24 per cent increase.

For the North East, North West and Yorkshire, the numbers the algorithm proposes are lower overall than the numbers delivered over recent years. But as with A-levels, the devil’s in the detail.

The really controversial changes are within regions, where the algorithm suggests jacking up numbers for shires, while taking them down in urban areas. Comparing the existing national formula to the proposal, we can see this for most large cities.

The number for Birmingham comes down 15 per cent, while the rest of the West Midlands goes up 52 per cent.

Numbers for Leicester go down 35 per cent. The rest of Leicestershire goes up 105 per cent.

Nottingham goes down 22 per cent, the rest of Nottinghamshire goes up 48 per cent.

Southampton goes down 17 per cent, Portsmouth down 15 per cent and Basingstoke down 23 per cent, but the rest of Hampshire would go up 39 per cent.

Wealthy Bristol would see some growth (5 per cent) but much lower than the rest of Gloucester, Somerset and Wiltshire (47 per cent).

It’s the same story up north. Leeds down 14 per cent, Sheffield down 19 per cent, and Bradford down 29 per cent. But the East Riding up 34 per cent, North Yorkshire up 80 per cent, and North East Lincolnshire up 123 per cent.

In the north west the core cities of Manchester (-37 per cent) and Liverpool (-26 per cent) see huge falls, while the areas around them shoot up. In Greater Manchester, for example, the growth is shifted to the blue suburbs and shires. Outer parts go up: Wigan up 10 per cent, Bury, up 12 per cent, and Rochdale up 97 per cent. And areas to the south and north of the conurbation up much further: Cheshire up 108 per cent, while Blackburn, Hyndburn, Burnley and the Ribble Valley together go up 149 per cent.

But it isn’t just that the numbers in the new formula are lower than the old formula for urban areas. In many cases the new formula suggests a lower number than their recent rate of delivery. This is true of Sheffield (12 per cent below actual delivery), Leeds (16 per cent), Bradford (23 per cent), the entire North East (28 per cent), Nottingham (30 per cent), Manchester, (31 per cent), Leicester, (32 per cent) and Liverpool (59 per cent). The new formula seems to assume we are going to level down our cities, not level up.

It’s true that there’s another step between the Housing Need Assessment which this algorithm produces and the final housing target, which can be reduced a bit to account for delivery constraints like greenbelt.

But if we go with this algorithm unamended, outside London most Conservative MPs will be seeing large increases in the housing targets for their constituencies, while many Labour MPs see their local targets reduced. Is this what we want?

Leaving aside the politics, I think not. Compared to the rest of Europe, the UK has much less dense cities.

Places like Dundee, Glasgow, Liverpool, Sunderland, Birkenhead, Hull and Newcastle all had smaller populations in 2017 than 1981, while places like Birmingham and Manchester weren’t much bigger. Our cities have untapped potential, many went through a period of shrinkage and have space, and there are health and environmental reasons to prefer urban growth too.

In dense urban areas, people are more likely to walk or cycle – and in the UK, people in cities walk twice as far as those in villages each year. This reduces public transport costs and improves health.

Denser cities can sustain better public transport and so cut car congestion and time spent travelling. As well as reducing pollution from transport, denser cities reduce energy use and pollution because flats and terraced homes are much more energy efficient.

I’m not sure the draft algorithm is even doing what Ministers wanted it to. The document in which it is set out says that “the Government has heard powerful representations that the current formula underestimates demand for housing in the growing cities in the Northern Powerhouse by being based on historic trends.”

But the algorithm seems to do the exact opposite.

There may be technical reasons why things aren’t working out: there’s lots of ways to measure affordability… differences between residence-based and workplace-based income measures… there were certain caps in the old model, population projections have changed and so on.

However, the bigger issue is this.

There’s no “objective” way of calculating how many homes are “needed” in an area. While there are ways of carving up the numbers that are seen as more or less fair, ultimately a vision is required.

Projections of population growth are circular: the projected population growth for the farmland between Bletchley and Stony Stratford would’ve been pretty low before we built Milton Keynes there.

Likewise the forecast for the derelict Docklands of the early 1980s. While there are real economic constraints, the future need not resemble the past.

Though it took a huge effort, Germany raised East Germans from 40 per cent to just 14 per cent per cent below the national average income since reunification. That’s levelling up.

Do we want to continue to concentrate growth in the South East? Do we want European-style denser cities, or for them to sprawl out a bit more? An algorithm can help deliver a vision: but it’s not the same as one.