Indices always have their critics. Take the Dow Jones, for instance. The decision was made, back in the day, to exclude IBM – and then to reintroduce it some years later. If this had not happened, the index would be double what it is now. Property indices are not much better. Take the last three years – a period where prices have been falling – or have they? It depends on who you talk to.

Both Savills and Knight Frank produce indices for London that show a fall in the market of between 16% and 8% since 2014. Over the same period, the Land Registry says it was down 9% in Prime Central London – while for the same areas, LonRes (a firm that aggregates market information) reckon they were down 7%. A big London Estate (a proxy for the market) had a valuation that was 4% up in 2015 and flat in 2016. These are big differences. Does it matter? We think it does; sentiment is a funny thing and often self-fulfilling. Buyers like buying into a rising market and a ‘falling market’ is a good reason to sit on your hands. It also matters in court, getting a mortgage or settling a divorce. These discrepancies, in what should be a transparent and sophisticated market, shouldn’t happen.

Why do they? Much of it is down to methodology. Savills and Knight Frank use agent valuation to produce their figures. A basket of flats and houses is assessed by individual offices once a quarter. Their indices have less to do with actual sales but a lot to do with how optimistic or pessimistic the valuer is feeling. When turnover is low, and your bonus is dependent on it, it’s easy to conflate turnover and value. Over time this can produce large discrepancies – so much so that one index had to be recalculated as it was wrong by nearly half. On the other hand you have the Land Registry which is compiled on completed sales. Unfortunately it is bald information that, for example, takes no account of lease length or size. It is also a lagging indicator as completions can be months – sometimes years – later. What you are measuring also makes a difference. Take so called Prime? Does it include Nine Elms? Or the Cromwell Road? You can start to see the problems.

So did we – and we are producing our own index for Prime Central London – but are setting about it in a way that we believe will answer the weaknesses outlined above. Perfection is impossible – but as accurate as possible will do. We have a database of market intelligence going back twenty-five years off which our London business is run. We have access to LonRes as a crosschecking tool and we know what we want – to produce information that is up to date, based on real sales in the streets and areas in which our clients would like to buy, that will stand up to forensic scrutiny and, ultimately, be the go-to index for a very specific area – Prime Central London. We have chosen to define this narrowly in post code terms as SW1/3/5/7 and 10, W1/2/8/10/11 and 14, NW1/3 and 8 and WC1 and 2. After much deliberation, we decided to exclude new developments being sold off-plan as the pricing information we have is not accurate enough; developers, for obvious reasons, prefer opacity to clarity.

Size, location, lease length and condition are the main variables in any property. Size is dealt with by using the price per square foot on exchange of contracts and adjusted for lease length. The house and the flat markets are different so we have split them. As mentioned earlier, we have rejected streets that are second rate in an area that would, as a whole, be considered Prime. We include, for example Onslow Square but reject the Cromwell Road. We compare properties in the same street but smooth out the data to make a meaningful index that tracks the change in the market. It is a method called Hedonic Regression and used by, amongst others, the Halifax Index.

The one factor that is very difficult to compress into an index is condition. The brains from the London School of Economics, who process the data for us, have not been able to crack this – but if the sample is large enough, this effect should not be statistically important. If you think about it, the number of houses or flats in really bad condition in Central London is tiny; hardly big enough to influence an average – an average being what we are after. We would be delighted to share the arcane details of the methodology – but you will need a wet towel over your head and an understanding of statistical maths.

What does it reveal?

Here are the graphs for houses and flats – more user friendly than raw numbers and better for illustrating trends. 

Putting this in percentage terms 2017 has seen a decline of 7.3% in houses and a plateauing of the flat market. Since the market peak of 2014, houses are off 20.7% and flats down by 14.7% – which demonstrates the efficacy of splitting the two. Houses are now below 2007 in real terms. One of the strengths of this methodology is that it is possible to parse the data to give an index by postcode, for instance – even down to different sides of the same square. We intend to update it twice a year.

What it shows is not pretty – and shows what you can do, as Chancellor, if you throw enough sand into the gear box. Unfortunately the current, politically weak, Chancellor decided in his last budget not to apply any oil. Despite that, we sense a change and suspect the lines on the graph will be flatter by the end of the year with an uptick in turnover. Not much – but better than 2017.