Property type, or property price?

This report has two themes. The primary them is to examine the relative importance of price (£/m2) compared with property type: flat, terraced house, and so on. In the second The secondary theme is the rebound in Prime Central London, which had been languishing at the bottom of the performance tables since mid-2016.

We look at a pan-London estimation of monthly repeat sales indices, specifically comparing the insights from classifying individual properties first by Land Registry property type, and for comparison classifying by the valuation quartile of the property’s postcode sector. The latter classification shows much stronger discrimination of significant relative price performance.

Land registry data for this report is updated to end January 2020. For our purposes we consider 8 postcode areas, two being central London and hence very small. The areas are N,E,SE,SW,W,NW,WC,EC.

Repeat sales records for the areas are pooled, and returns estimated by the method described elsewhere. The same records are partitioned two ways, in each case into four subsets.

  1. According to the Land Registry type: Flat, Terraced, Semi-detached, Detached. This is highly unbalanced, reflecting the preponderance of flats and terraces.
  2. By estimated aggregate present value per square metre of the postcode sector. The quartiles are designed to have the same number of single-sale properties with floor area. The total number of repeat sales is also reported.

In both cases the input data are disjoint (there is no overlap) and no post-filtering either spatial or temporal has been applied so the indices are independent in this sense. The repeat sales input data are seasonally adjusted.

Results by property type
subset N1 m2/unit m2 change2 n.repeat3 R2 £000/unit £/m2
Flat 798 66 0.2 475,900 0.93 590 8,850
Terraced 344 106 2.1 195,300 0.95 890 8,350
Semi-detached 90 128 2.7 47,400 0.95 1,030 7,990
Detached 21 197 7.6 10,700 0.93 1,900 9,630

1 tracked units (000)

2 mean m2 change between first and last measure (individual units)

3 repeat sales count

Results by \(£/m^2\) quartile
subset N m2/unit m2 change n.repeat R2 £000/unit £/m2
Q4 (£ High) 322 94 1.5 202,100 0.94 1,210 12,870
Q3 325 83 1.2 197,600 0.95 700 8,410
Q2 305 79 0.8 164,300 0.95 540 6,790
Q1 (£ Low) 301 80 0.7 163,100 0.95 420 5,280

The results show that \(£/m^2\) has significant power to discriminate relative short-run performance as described in previous white papers. Specifically over the 3-year period mid 2016 to mid 2019 the top quartile returned nearly -10% and the bottom quartile +6% in log terms, corresponding to an arithmetic underperformance of the prime sector of 17%. As the top quartile is valued at £388Bn and the total tracked London market at over £900Bn, the amounts involved are quite economically significant.

Property type is of much lesser usefulness than \(£/m^2\) because it is not the relevant metric - some of the highest quality / highest priced residential real estate is in the form of apartments, but some of the lowest quality ex-local-authority stock is also in this bucket, and flats are also most common in the central districts. This lack of price discrimination is quantified in the tabulation by property type, \(£/m^2\) column. In addition there is the serious signal-to-noise problem which is a direct consequence of the small number of detached and semi-detached houses.

A noteworthy point: individual flats’ floor area growth averages to nearly zero, whilst houses - and especially detached houses - demonstrate an ability to grow. This is of course common sense, and it’s somewhat reassuring that we see these sensible results at least at the aggregate level despite the noise in individual EPC measurements. For a more controlled metric this ‘growth’ phenomenon could be accrued into time buckets using the same technologies applied to performance index estimation.

We might be surprised that despite significant short-run divergences there is not a greater disparity of long-run performance such as we see at the national level, and that in fact the middle quartiles have outperformed by a very small amount the top and bottom quartiles, so split at the median the two sets of returns are very similar. Some important components of total return are missing however: net capital expenditure, service charge/ground rent where relevant, and of course rental yield if it is an investment property. To some extent some of these may offset one another (in particular maintenance capex trades off against service charge) but we don’t have data to further investigate these aspects.

There is ‘by inspection’ a lead/lag relation between the series, with high-price quartile 4 leading the big upswings. This requires careful specification however if it is to show up in test statistics. To illustrate this point I show two ‘shocks’: the first is the financial crisis, the second is the 2019 PCL reversal discussed here. In the crisis the timing and magnitude of the fall was uncannily indistinguishable between quantiles, especially the date of onset and reversal. Last year 3 out of 4 quantiles were down in May, and all were up in June (recall that these are seasonally adjusted and the May/June seasonals are anyway very similar. Whilst the onset was simultaneous, the magnitude over 6 months has strictly followed price rank ordering - this will require careful specification of the model possibly using copulas but this topic goes a little deep and will not be explored further here.

Up until May 2019 Prime Central London was falling quite fast, approximately 4% over the preceding 12 months and an aggregate 9% since 2016. Price convergence was occurring across the spectrum, with lower priced areas consistently outperforming.

All this abruptly reversed in June. Whatever your views, the following facts stand out.

  • 2019-05-16 Boris Johnson confirms that he will run for the Conservative Party leadership after Theresa May stands down.

  • 2019-05-24 Prime Minister Theresa May announces her resignation as Conservative Party leader, effective 7 June.

The prospect of a leadership change in the ruling party galvanised domestic and overseas buyers alike. The latter having been waiting for some resolution of the political chaos snapped up ‘bargains’ in the crucial Prime Central London market after 3 years of decline and a sterling devaluation, seeing an increased likelihood of an investor-friendly majority in parliament and a possible sterling resurgence. This was vindicated 6 months later, and the rally has been sustained thus far into 2020 with a 7% rebound corresponding to a £25Bn book capital gain in the top quartile.

It is especially noteworthy from a technical perspective that the estimation method’s ability to resolve an abrupt V-shaped reversal with a clear ‘smoking gun’ chain of causation is here vindicated, so the criticism of ‘all property indexes show artificially smoothed returns’ is strongly refuted.

Editor’s note: at the time of writing, Covid-19 was not yet a talking point.

Giles Heywood
price paid last transaction: 2020-01-31
report run: 2020-07-05

This study was carried out on request to investigate the dynamics between property types in the London market