Assessing House Prices: Insights from HouseLev, a Dataset of Price Level Estimates

Mardi | 2018-12-04
Salle des thèses 16h – 17h20

Jean-Charles BRICONGNE – Alessandro TURRINI – Peter PONTUCH

Despite growing consensus on the relevance of a sound assessment of housing market developments for macro-financial surveillance, housing prices assessments in a multi-country context normally rely on price indexes only. This has a number of limitations, especially if available time series are short and series averages cannot be taken as reliable benchmarks. To address this issue, the present paper computes housing prices in levels for all EU countries and a number of other advanced and emerging economies according to common methodologies. The baseline methodology makes use of transaction data or on information on the total value of dwellings in national accounts statistics and on floor areas of existing dwelling stocks from census statistics. When such information is not available, price level estimates are based either on house price offers quoted in realtors’ websites. When both sources are available, discrepancies usually do not exceed a few percent, which confirms the validity of this second approach despite an expected upward bias. House price level estimates permit to compute price to income ratios with a clear interpretation: the average number of yearly incomes necessary to buy dwellings 100 m2 large. Using a signalling approach aimed at identifying price-to-income threshold maximising the signal power in predicting downward price adjustments, it is found that a price to income close to 10 can be taken as an across-the board rule of thumb for identifying possibly overvalued house prices. Besides, when price levels are used in regression-based models to estimate fundamental-based house price benchmarks, they permit to exploit cross-section variation thereby providing additional insights as compared with analogous benchmarks based on house price indexes.