STATEC

Unit SOC4: Price statistics

13, rue Erasme L-1468 Luxembourg

The indices measure changes of the acquisition price of new and existing apartments and existing family houses in Luxembourg over time. They are based on the official prices as indicated in notary acts which have been registered at the 'Administration de l’Enregistrement et des Domaines' (AED). Furthermore, the indices measure pure price changes as differences in quality of apartments dwellings sold at different times are neutralised (hedonic price indices).

A distinction is performed between 'Ventes' (VEN) and 'Ventes en état de futur achèvement' (VFA). This is an administrative distinction based on the legal type of transaction. VFA are transactions regulated by a specific law (Loi du 29 décembre 1976 relative aux ventes d’immeubles à construire et à l’obligation de garantie en raison de vices de construction). VFA are always 'new dwellings' whereas VEN are typically 'existing dwellings'. From 2015, a further breakdown is performed for 'new dwellings" according to the purpose of the acquisition. Two sub-indices are compiled: IOOH, accounting with the transactions for owner-occupiers purpose and IINV which includes the dwellings purchased as investment.

HPI indices cover all residential properties purchased by households in the reference period.

Price indices are based on the acquisition principle. Therefore a transaction enters the index on the date indicated on the notary act. All registered transactions are considered.

Private households.

The target population is the set of all transactions of dwellings purchased by households.

Luxembourg.

Price indices are available from the 1st quarter of 2007 onwards.

Price indices are published in terms of a base set at 100 on average in 2015.

The following units are used:
• Indices;
• Rate of change (quarterly and annually).

Quarter.

Regulation (EU) 2016/792 of the European Parliament and of the Council of 11 May 2016 on harmonised indices of consumer prices and the house price index, and repealing Council Regulation (EC) No 2494/95

not applicable.

Data are published up to the level of detail described in the HPI Regulation.

Only aggregate indices are published.

Usually the data is published 85 days after the reference quarter.

The release calendar is available at the Statistics Portal.

HPI indices are published on the Statistics Portal.

The indices are published quarterly.

The data is published as a short-term indicator available of STATEC's website: Indicateurs à court-terme - Statistiques - Luxembourg (public.lu)

 (set C)

”Le logement en chiffres” is a joint publication of STATEC and Housing Observatory (which works in collaboration with LISER). This bi-annual publication brings together the latest available data on housing sales (available in French): 

Data is published as an Excel file available on STATEC’s website.

The publication appears as 'Indicateurs à court terme - Série C'. This data is also released by Eurostat.

HPI indices are published on the Statistics Portal (Excel file):

https://statistiques.public.lu/dam-assets/fr/donnees-autres-formats/indicateurs-court-terme/economie-totale-prix/D4011.xls

and

https://statistiques.public.lu/dam-assets/fr/donnees-autres-formats/economie-totale-prix/prix/d4016.xlsx

Micro-data is not published.

See 10.1

Eurostat Technical manual on Owner-Occupied Housing available online.

Methodological details of the price indices are published on STATEC’s website: Prix des logements - Statistiques - Luxembourg (public.lu)

Several documents are published on the Statistics port:

Note méthodologique sur la statistique des prix de vente des appartements
Economie et statistiques N°79/2015 - L'indice des prix des maisons anciennes
Economie et statistiques N°44/2010 - Un indice des prix hédonique des appartements

Residential Property Prices - Statistics Portal - Luxembourg (public.lu)

Micro-data on which the index is based are checked for outliers.

The quality of the HPI index is routinely reviewed using a framework that is based on the OOH technical manual and the European Statistical System (ESS) definition of quality.

The HPI responds to the needs of all users interested in knowing the behaviour of prices in the housing market. These price indices are based on registered transaction prices.

No user survey has been done yet.

The series cover transactions of new and existing apartments as well as existing family houses (since the 4th quarter of 2012) in the whole country.

The overall accuracy of the price indices is good. The price corresponds to the official registered price.

None. The indices are based on an exhaustive data base ('Publicité Foncière de l’Administration de l’Enregistrement et des Domaines'). As additional transactions become available from previous quarters, they are included. Hence revisions are possible.

Not applicable.

Data is published at the latest t+90 days after the end of the reference quarter.

Data should be sent at latest on the target day. A delay can occur due to unforeseen circumstances (sickness of people calculating the index, technical problems with the production system).

Price indices are aimed to be comparable with other house price indices developed in the European Statistical System in the context of the Owner-occupied housing regulation.

The data for apartments are fully comparable over time. The same for single-family houses along the period covered (since the 4th quarter of 2012). However, the integration of houses introduced a break in the aggregated series, which previously only included apartments.

The HPI partly overlaps the scope of the other owner-occupied house price indices described in Commission Implementing Regulation (EU) 2023/1470.

Five price indices are published:

1.    A price index for new apartments;
2.    A price index for existing apartments;
3.    A price index for existing single-family houses;
4.    A price index for existing dwellings (aggregation of 2 and 3);
5.    A price index for all dwellings (aggregation of 1, 2 and 3).

Around 7 days per quarter is needed for the regular production and validation of the results, time of updating and improving the production system not included.

The HPI is subjected to revisions according to the Commission Implementing Regulation (EU) 2023/1470

As additional transactions become available from previous quarters, they are included. Hence revisions are possible. Data may also be revised because methodological improvements have been performed.

See below.

The price indices are based on a database managed by the AED called 'Publicité Foncière'. This database covers the information contained in the notary acts which must be sent to the AED for every real estate transaction.

The weighting scheme is also derived from the 'Publicité Foncière' (administrative data file).

The database used for the HPI is also used for House Sales Indicator.

During month t, a data file is provided to STATEC with the transactions recorded at the AED during month t-1 and which covers notary acts signed during month t-1 and month t-2.

The data is contained in a XML file which is transferred from the AED to STATEC. An additional survey is run among households who purchased existing houses in order to collect characteristics of these houses, such as size or type of house.

During the production process, a number of data checks are performed. The different sub-steps of the pre-processing algorithm are monitored. The resulting indices are benchmarked against non-adjusted average prices. Statistical indicators are compiled for different strata. These indicators are not published but are only used for validation purposes.

See below.

See below.

HPI is a Laspeyres-type price index.

The hedonic method is combined with a stratification that distinguishes between new and old apartments.In this way, it is possible to obtain a separate index series for these two market segments.The two sub-series are then linearly aggregated to obtain a price index for total apartments that is consistent with the two sub-indices.

Since 2012, a sub-indice for existing houses is included in the HPI.

Weights are based on household expenditure on transacted dwellings.

Weigths are updated every year.

The weight reference period is t-1, elementary aggregrate weights are price updated between year t-1 and Q4 of year t-1.

The weight reference period is t-1, elementary aggregate weights are price updated between year t-1 and Q4 of year t-1.

The sub-index series (existing houses, existing apartments, new apartments) in base 100 equal to the 4th quarter of the previous year are linked with the corresponding long series (currently in base 100 in 2015) taking as a link the 4th quarter of the previous year. The chained indices are published.

An hedonic model is used for all sub indices: newly built apartments (since 2015 sub divided into apartments for owner occupied / for investments. These sub indices are not published), existing apartments, existing houses.

The hedonic repricing model function explains the price of an apartment according to its characteristics. This econometric relationship constitutes the basic structure underlying any hedonic index. After many tests on the prices and characteristics of the apartments provided in the AED database, the following hedonic function finally led to the most satisfactory results.

The logarithm of the transaction price is explained in terms of the following variables:

• The surface of the apartment
• The square root of the surface of the apartment
• The type of sale (Old=1; New=0)
• The existence of a cellar (Yes=1; No=0)
• The existence of a garage, car park or location (Yes=1; No=0)
• The distance (in km) between the municipality where the apartment is located and the City of Luxembourg

For existing houses, the following characteristics are taken into account:

• logarithm of surface
• logarithm of the land area of the plot
• logarithm of the distance (in km) between the municipality where the house is located and the City of Luxembourg
• inverse of population density
• type of house (detached, semi-detached, detached)
• house age
• number of bathrooms
• number of covered parking spaces
• carrying out major works

not applicable

The hedonic method is combined with a stratification that distinguishes between new and old apartments. In this way, it is possible to obtain a separate index series for these two market segments. The two sub-series are then linearly aggregated to obtain a price index for total apartments that is consistent with the two sub-indices. New apartments are sub divided since 2015 into apartments for owner occupied and for investment.

See below.

The day taken into account for pricing is that of the notary act, which is included in the dataset. Each transaction is then allocated to the quarter including that day.

VAT is included in the price of new dwellings.

Taxes other than VAT are not included in the price of dwellings.

Government subsidies are not included in the HPI.

The amount taken into account is the full, total transaction price, which includes the land component.

Housing cooperatives are not covered in the HPI.

Non-market transactions are not included in the HPI, nor in the House Sales Indicator, as long as they are identifiable as non-market transactions.

Multi-object transactions are not included in the calculation of the HPI and not in the House Sales indicator.

Fractional transactions are not included in the calculation of the HPI, nor in the House Sales Indicator.

Apartments: 

All transactions that do not include at least one private lot whose description is similar to an apartment and for which the surface area is not mentioned are rejected. As the statistical interest essentially relates to prices, transactions for which the amount does not appear in the file are not considered. In order to limit ourselves to typical apartment sales, transactions that do not meet all of the following criteria are eliminated:

• The transaction is either an ordinary sale or a sale in future state of completion.
• The transaction must involve a transfer of a right of ownership.
• The quota must be equal to 100%.
• The number of different proprietary lots making up a transaction is less than 13.
• The surface of the apartment must be less than 300 m2.

Before the compilation of the statistical series, an automatic processing is applied in order to detect the observations which can be qualified as extreme compared to the majority of the observations. Such transactions may indeed correspond to too specific real estate or to input errors. Two steps then take place, the first being based on quartiles and the second on an econometric model.

First, a price per unit area is calculated for each observation. For each quarter, the transactions are grouped into classes according to the surface area of the apartment and the type of sale (new or old). Within each class, the first quartile q1 and the third quartile q3 are calculated on the normalized prices. Finally, an observation is considered extreme with respect to its class if its normalized price pn is such that:

pn ≤ q1-1.5∙(q3 - q1) or pn ≥ q3+1.5∙(q3 - q1)

After this first cleaning, a statistical model is estimated explaining the transaction price of an apartment according to its characteristics. A transaction is then considered extreme when the difference between the estimated price and the price actually observed exceeds a certain threshold.

Houses: 

The transaction is excluded if number of bathrooms > 4, or covered parking spaces > 4. A second stage of filtering is carried out by means of 2 regressions. For each transaction, these regressions respectively link the logarithm of the price to different explanatory variables:

• The logarithm of the energy reference surface (first regression), in order to eliminate transactions that are atypical due to the building
• The logarithms of the capacity, the distance in Luxembourg as well as the inverse of the population density (second regression), in order to eliminate transactions that are atypical due to the terrain

We proceed in this way, with 2 separate regressions rather than just one, in order to avoid compensation phenomena between the surface of the house and the capacity of the land. Thus, if a house is very expensive, or on the contrary very cheap given its surface area (1st regression) or the capacity of its land and its location (2nd regression), the residual of the corresponding regression will be very large in absolute value for this transaction.

Not applicable

No adjustments other than quality adjustment are applied to the HPI data.

None.