UK Housebuilders Completions and Productivity

In this blogpost I hope to provide evidence and data to show that the UK's top housebuilders as churning out houses as fast as they can, rather than holding back to preserve profit margins.

The Biggest Housebuilders

The UK's largest housebuilders in terms of completions, that is houses built in the years ending March 2016 and March 2017 are as follows:-

	2016	2017
Barratt Development Plc	17,319	17,395
Persimmon Plc	15,171	15,588
Taylor Wimpey Plc	13,808	14,112
Bellway Plc	8,721	9,644
Redrow Plc	4,716	4,918
Bovis Home Group Plc	3,977	3,755
The Berkeley Group Holdings Plc	3,776	3,905
Galliford Try Plc (Linden)	3,078	3,296
Crest Nicholson Plc	2,870
Countryside	2,657
Quadrant Construction	2,510	2,552
Bloor Homes Limited	2,443
Keepmoat	2,416
Miller	2,380
Avant	1,210	1,636
Nottinghill Housing	1,170	1,151
Morris	1,165
CALA	1,151
PfP	1,119
Guinness Partnership	908

These figures are taken from the Annual Reports of each of the companies. The government's Department of Communities and Local Authorities records that the total number of home completions for 2016 was 139,840 and for 2017 it was approximately 153,000.

Therefore the top 20 largest housebuilders account for between 50% and 70% of all completions in the UK.

Past Decade Performance

By looking at previous Annual Reports we can build up a picture of year on year performance, since 2004, including the 2008 credit crunch.

The data is sparser the further back we go because I am lazy and cannot spend so much time looking for historic company reports.

A few observations:-

• We can see that most of the companies have returned to approximately the 2008 levels.
• 2007 and 2008 seem to be exceptional years rather than the result of steady year on year increases.

The year on year increase since 2008 is on average for all the companies 10% each year, however, there is some degree of bumpiness

Cost of Building

From studying the Annual Reports we can estimate the cost of building a house. If the total revenue for a company is X and the operational profit is Y, then X minus Y is the total cost of all the labour and materials and equipment. This ignores any admin expenses, back office overheads and interest on loans. By dividing total costs by the number of completions, we get an approximate cost per house.

For the top 14 companies the cost per house is as follows (using most recent data)

	Cost per House
Barratt Development Plc	£221,385.46
Persimmon Plc	£155,902.71
Taylor Wimpey Plc	£210,884.99
Bellway Plc	£181,325.18
Redrow Plc	£248,338.26
Bovis Home Group Plc	£226,326.38
The Berkeley Group Holdings Plc	£456,773.37
Galliford Try Plc (Linden)	£232,736.65
Crest Nicholson Plc	£276,376.31
Countryside	£254,001.39
Quadrant Construction	£126,895.77
Bloor Homes Limited	£249,662.71
Keepmoat	£441,556.29
Miller	£194,117.65

We can see that the range is from £126,000 for Quadrant Construction to £456,000 for Berkeley. However, the average cost per home is around £222,800 (this is for 61% of the homes built).

We can speculate that these companies aren't trying to get build as efficiently as possible because their profit margins will easily cover expenses, and so it would be possible to build houses by far cheaper means, but this is literally the majority of the market accounted for.

The crushing inevitability of MiData

Yesterday there were news stories about how the average UK consumer could save up £70 a year by changing bank accounts. Unbeknownst to me, the government had launched an initiative months ago, urging banks to allow customers to download their bank account transaction history in a standard format, namely MiData.

MiData is a comma separated value text file, you can open it in NotePad or Excel.

At the moment pretty much the only two things you can do with MiData is faff about with it in a spreadsheet, or upload it to GoCompare who will somehow process it and tell you which bank to change to.

I think GoCompare just looks how far into your overdraft you go and what the average account balance is, they then look at which bank accounts charge and pay what interest and other goodies and make recommendations. Their best recommendation for me was some Yorkshire bank who charge higher interest, but give you a £150 switching bonus, so less of a saving, more like a one-off free gift.

Anyhoo, there's so much more potential and risks involved with MiData.

Years ago I read online, possibly from Worstall, of an idea for banks (with the user's permission) to mine your data and automatically save you money by changing various service providers. For example say your current energy provider charges £30 a month, but other people in your area with the same household size are only paying £20 with a different provider, then the bank would change you over, saving you £10 a month. Presumably the bank would pocket half your saving for a limited period, but since you're paying less, who cares. No bank has done this, probably because of privacy laws.

With MiData, the ability to minedata is outwith your bank. But at the moment, there are no tools, no services. The main risk is that the MiData is just too personal.

When your bank lets you download the MiData, it is "anonymised" which by the looks of things means they remove any account numbers, and anything that look like an account number, just replacing it with asterisks. This only makes it anonymous in that you don't know personal account details, but that's not enough.

As an aside, I understand that some car insurers fit a black box that records your car's speed and time, so that they can insure you appropriately for how safely you drive. I read that some researchers can use this speed data alone to figure out where you are going each day. It takes a bit of datamunging, but presumably if you know the start point and the junction one way is 30 seconds drive and the junction the other way is 50 seconds drive. Any nefarious criminal can map your life just from speed measurements.

Similarly, from MiData, even without account details, it would be trivial to identify a person from their transactions.

For example, looking at petrol stations and supermarkets you can get a feel of where in the UK a person lives and works, they'd do their weekly shop within one or two miles of their house, their regular petrol fill up will be somewhere between their home and their place of work. Or even better their local train station or work train station will be within less than a mile. Occasionally they will be travel or petrol transactions further away, these would be holidays or visiting family members, traditionally some family members stay in the same place where they grew up. Likewise gift purchases will coincide with birthdays. An investigator can get themselves to Linkedin and Facebook and look for people who live in this area, work in another area and grew up some other specific place, and who's partner / parents have birthdays at whatever time of year.

There aren't many people who live in Chingford and work in Hertford, even fewer who grew up in Manchester.

Anyhoo, the cat is out of the bag. Like in the book The Light of Other Days by Stephen Baxter and Arthur C. Clarke, the post-millenial generation aren't going to give a crap about privacy, compared to the "benefits" of datamining. I imagine that security expert Bruce Schneier would be doing his nut in.

So, having identified a gap in the market, I have an awesome idea for a business that will turn me into the millionaire I've always dreamed of being.

First we create an app or website where people upload their MiData to and the site gives you a neat pie chart showing how you spend your money in categories like supermarket, petrol, Entertainment, etc, and histograms showing how much you spend on each category each month. Just like Quicken used to do before they discontinued the UK version.

Don't worry, your data has already been anonymised by the bank, the government said so.

Then once we have enough people's "anonymised" data, we add some data, like geographic locations for each supermarket, train station and petrol station and cafe, then we offer website users a fancy map showing where they spend. People will think its ace, and Bruce Schneier will start getting worried.

Then we do some more analysis showing how much people in different areas are spending on things, like the aforementioned energy providers, and we can start charging users for recommendations for where to switch to.

Then we can start telling people how many kids we think they have based on their data, and how many bedrooms their house has, recommendations of which car they should buy next, which phone and whether they are engaged in illegal activity, or what things they do that are abnormal.

The problem is that I don't have time to do this, neither have I the skills. Someone else will.

The government will at the same time as encouraging it and providing grants to organisations who can take advantage of the MiData, will also have very legitimate concerns about privacy.

There is a very faint trend on social media for young people, teenagers who have just received their first ever credit card, to post photos of said card and unwittingly give away the security number, so that nefarious people will use their details. Young people can be stupid. Lots of people are stupid and will do stupid things.

The government, and parents too, have a difficult job in weighing up the benefits of things like MiData and credit cards, with the risks. What protective measures will they put in place that are just as much of a ballache as the EU Cookie Directive, that makes you have to click on disclaimers on websites.

Imagine, if you will legislation that protects people's MiData privacy by putting in place some hardcore digital rights management, only allowing special government approved organisations and businesses to view and process, thus no small app developer could play with the data, only GoCompare and the banks and probably government departments, HM Revenue & Customs, and the police, probably hospitals too. Some DRM system that's so encrypted and heavyweight that developers often do raw datadumps, and leave hard disks and DVDs on trains.

This is what happens.

Getting all Fifth Normal Form on your ass

I've been reading up on database normalisation and thinking about how the London Indiepop Eyespy website would work.

Data normalisation is incredibly interesting, but not quite as important as what the site looks like, form dictates function here.

In the original pub game of indie eyespy, you'd spot a band member then look up how many point they are worth.

There's Stuart Murdoch from Belle and Sebastian, four points!

As many of the bands are quite obscure, it would be less likely that you'd know the names of band members, but you could still get points for identifying the person.
There's wassisname the drummer from Camera Obscura, would be worth the same as correctly remembering his name was Lee. Although, if playing indie eyespy competitively with friends, the person who get his name would claim the points.

So here we have the dilema for a website, what should it list:-

• Check boxes for each band member by name for each band
• Check boxes for numbered members for each band
• Text box for number of members spotted and leave the rest to banter

With the limitless possibilities of computing power, I'm tempted to go for option one, and make life easier for folk who can't remember names

As bands evolved and lineups changed so did the way points were allocated, for example spotting Gav from Camera Obscura would get you three points, but if you instead noted it was Gav the bass player from Stabiliser, it would be five points. However, you couldn't then claim eight points for identifying him as playing in both bands. You only get points for one person once.

So the website would need a way of dealing with the same person in many bands.

With the Skilmo website, this sort of problem was easily resolved when it processed the checklist page, each skill was only logged once, for example tunisian crochet only counted once if you clicked it in both the textile arts category and the crochet category because the program ignored categories when logging stuff.

But with indie eyespy, because of the different points values this isn't possible, and rather awesomely, this is precisely in what 5th Normal Form is about in database normalisation theory.

On the database side of things I'd need three tables:-

• List of people and their nominal point values
• List of bands and their points values
• List of which people are in which bands

Then every time the website is accessed a list is constructed by SQL of bands and their members and points values.

Then with PHP an html page is generated which displays this list, with form checkboxes and also generates so neat JavaScript which greys out people who play in different bands when you select them.

Then submitting this form should post a list of people and their points values.

The next page would get a list of the people in the database, and run through the posted data, tote up the score, and log the score.

That seems to work in my head.

Of course the list of names will be normalised with id numbers to cover for there being more than one person with the same name, and same with the list of bands.

The list of bands also needs a status column for whether a band is currently active, on haitus or split up. Cos, of course, active bands are worth more points than bands that are on haitus, but not as much as a band who split up in 1989.

And the list of which people are in which bands needs a column indicating former band members. Because, as I'm sure you understand, spotting the original guitarist from Pocketbooks is worth more points than the current guitarist.