On the "best" evidence (see below), the average UK house price is only 3% below the 2007-08 peak. The average home-owner, however, has experienced a decline of 11%. The difference is explained by the outperformance of expensive properties.
 
The wide range of house price measures showing divergent performance is a boon for biased commentators seeking evidence to support their particular thesis. The chart compares five widely-quoted measures, rebasing each to 100 at the 2007-08 peak*. All five claim to be constant-quality measures and have been adjusted for seasonal variation. (Four of the five providers publish seasonally-adjusted series; the DCLG index has been adjusted within Datastream.)
 
Bears, naturally, focus on the Halifax / HBOS index, which is down by 20% from a peak reached in August 2007 and fell by 4% in the year to May. At the opposite end of the range, the LSL Academetrics index is 3% below its February 2008 high and has risen by 1% over the last year.

The construction of the various indices differs in numerous respects but a key distinction is whether a particular gauge aims to measure changes in the average price of property or in the price of an average property. In the former case, price changes are combined using value weights, thereby assigning greater importance to more expensive properties. This is appropriate if the focus of interest is movements in the aggregate value of the existing housing stock. Volume weighting, by contrast, treats expensive and inexpensive properties identically, resulting in an index that measures the experience of the average home-owner. (In stock market terms, this is equivalent to the distinction between share price indices measuring performance using market capitalisation and equal weights.)
 
LSL Academetrics and DCLG use value weights (average property price) while the Land Registry (LR), HBOS and Nationwide use volume weights (average home-owner experience).
 
Other important differences include: 

• Sample representation: The DCLG, HBOS and Nationwide measures are based on mortgage transactions while the LR database – also used by LSL Academetrics – includes cash sales.

• Sample size: The LSL Academetrics index is based on the largest sample, followed by DCLG and LR. HBOS and Nationwide numbers vary with their share of the new mortgage market.

• Stage of purchase process: HBOS and Nationwide record prices on mortgage approval and DCLG on mortgage completion. The LR and LSL Acadametrics indices use prices registered with the Land Registry, implying a typical 2-3 month lag relative to HBOS / Nationwide.

• Quality adjustment: LSL Academetrics and DCLG base their indices on a constant mix of properties while HBOS and Nationwide use hedonic regressions to correct for quality difference. The LR index is based on repeat sales (like the US Case-Shiller index). These different methodologies should, in theory, produce similar results.

These considerations suggest that the LSL Academetrics index is the "best" value-weighted measure, since it is based on the largest sample and includes cash sales. The LR index is the preferred volume-weighted measure for the same reasons. The trade-off is that they are less timely than the HBOS and Nationwide indices.

As noted, the LSL Acadametrics index was 3% below peak in May. The alternative value-weighted measure, the DCLG index, was 5% lower in April (the latest available month). The difference may be explained by the former's inclusion of cash sales, assuming that these have been associated with stronger prices. The LSL Acadametrics measure rose by 1% in the year to May versus a flat performance of the DCLG index in the 12 months to April.

The LR index, measuring the experience of the average home-owner, is 11% below peak and fell by 1% in the year to April. Interestingly, this accords with the Nationwide measure – 11% below peak and down 1% in the year to May – but not the HBOS index – 20% and 4% respectively. The recent significant divergence of the HBOS index from other measures suggests that it should be given less weight in assessing market trends.

House prices remain elevated relative to incomes but not rents. A measure of the rental yield derived from the national accounts (i.e. the sum of actual and imputed rents divided by the value of the housing stock) is exactly in line with its average since the early 1960s – see second chart and previous post for more discussion.
 
Forced selling pushed house prices to an undervalued level relative to rents in the mid 1970s, early 1980s and early 1990s but is much less of a factor today, reflecting low interest rates and a stable labour market. Rents, moreover, are growing strongly as tenant demand outstrips supply, so an increase in the rental yield to above its long-run average could be achieved without any further fall in house prices.

*Other measures include the Rightmove and Hometrack indices, both based on surveys of estate agents. The Rightmove index uses asking rather than sale prices while the Hometrack measure is based on a relatively small sample size.