Skip to content

Statistical processing

Contact info

Short Term Statistics, Business Statistics
Paul Lubson
+45 39 17 35 42

pal@dst.dk

Get as PDF

Construction

The construction statistics are compiled on the basis of data extracted from the Central Register of Buildings and Dwellings and is updated quarterly. Data valuations (correct errors) and a number of controls (valuations of quality) are carried out. Furthermore, different diverted variables and delimitations are formed. There are both non-corrected figures ("raw" figures) and corrected figures (because of delayed reporting). The quarterly publications also contain a seasonally adjustment.

Source data

For updating a building statistics register Statistics Denmark receives monthly data from the Central Register of Buildings and Dwellings.

Frequency of data collection

Data are collected monthly, but published quarterly.

Data collection

The construction statistic are based on the reports from the municipalities concerning the actual activity in building construction to Building and Dwelling Register (BBR). When the municipalities issued a building permit, it is reported to BBR. Statistics Denmark receives every month a copy of BBR. Every quarter, the monthly deliveries in the quarter are added up and then the register of building cases (BSR) is updated - when the data have been "cleaned", filed and the derived/compounded data are created.

Data validation

In connection with the import of figures to the Register for Building Cases (BSR) is there carried out a process of correcting errors where the following buildings cases are removed:

  • Duplicates where the building_Id, date for building project and litra are identical.
  • Building cases which are registered as ended without being completed. These cases are when the building has changed the number of property in case of dividing up or the like.
  • Building cases which have no date of building project. It is buildings which are purely registered because of § 20 in the BBR-law. This is done because it should be possible to see future building projects when the building is traded.
  • Building cases with negative intake of dwellings and/or area, typically demolitions.
  • Temporary buildings. In connection with the publishing of 2nd quarter 2015, there is made a change of method where temporary buildings, typically pavillons and site huts, is removed back to 1st quarter 2010 and forward

The control consist of basically in four elements, namely control of if there are reports for all months and all municipalities in the reference quarter, if the reported "raw figures" (non corrected figures from the municipalities) contain errors, if the estimations (correction for delayed reporting) and the seasonal adjustment looks reasonable.

The updated BSR (register for building cases) is subject to a number of controls:

  1. Figure import control: In this control there is a control partly of, if the date of delivery is correct (shall be in the actual delivery month), and partly of, if the parts of imported elements are on a satisfactory level, e.g. there shall be addresses on all buildings and there shall be a land register attached to all buildings
  2. The total overview: It is a making up of permitted, started and completed construction, calculated on number of dwellings, total floor area and number of building cases. They are undertaking a valuation of, if the figures looks reasonable. They shall be higher than last quarter. Simultaneous, the Model version and the timetable is controlled.
  3. Logical control for "outlayers", which exceed defined limits for number of dwellings, average size of dwellings and total floor space, checked against the use of the building and phase of construction
  4. Individual control: This is a manual control of probability where focus is on the construction distributed on municipalities. The individual municipalities reports of the total construction for permitted, started and completed buildings is compared with the similar quarter 3 years back. If there are "suspicious" developments, all the building cases from the municipality concerned looking through and the "suspicious" building case is identified and analyzed. If it is evaluated as an error, the municipality is contacted. Possible errors are corrected in BSR.
  5. Sum control: In connection with the estimation calculations there is partly a control of, if all the new months are included, and partly of, if the estimated figures looks reasonable.
  6. Estimation control: The control consist of a evaluation, how good the estimations on permitted, started, under construction and completed construction hit the objective on +/- 5 p.c of the "final" figure (7. estimat).
  7. Control sheet: In connection to the seasonal adjustment there is a control of, if all months in the reference quarter are there, and if the figures looks reasonable, e.g. if they are following the general tendency.

Data compilation

When the "raw" figures (not adjusted for delays) are enclosed in the BSR, an updating of the buildings address is carried out. Delayed reports from the municipalities are the construction statistics greatest problem.

Statistics Denmark has since 1996 established models for estimation, which corrected the "raw" figures for these delays. In connection with the publication of 1. Quarter 2019 employed (the former model was from 2017). The model estimates the extent of the delays and corrected the reported figures for this in the last 18 months. The greatest challenge in the project has been to find a permanent pattern in the delays. The definition of a delay is, when a reported building case does not concern the actual counting month, so it is delayed. In http://www.Statbank.dk the changes between the first reported figures and the current figures are published by phase of construction in table BYGVFORS.

The latest model for estimation is from 2012. It is based on the original model from 1996. The starting point for the model is, at analyses showed, that after 18 months there are very small revisions in the reported figures. Therefore, these are regarded as "final", but are still revised. Analyses also showed that the phase of construction, use of buildings, type of building case and type of builders were good variables for explanations. At the same time, it was clear, that it was not possible to make estimations on municipality level. There were too few data in particular the small municipalities.

The new model for estimation is building on the former model which takes into account:

  • The state of the building permit, e.g. permitted or completed building.
  • Floor area or number of dwellings, which have different pattern of delays.
  • The classification: Use of Building (15 main groups) .
  • New building or conversion/extension of buildings, etc.
  • The classification: Ownership (private, non-profit building society or public)
  • An ongoing revision at the times of publication, taking into account the delays observed. For each state of building project a raising-factor is estimated using analyses of regression for delays.
  • No estimate is made for the geography and the number of buildings.

The model: The groups of variables gives total 90 different groups to consider for every of the 8 phases of construction. On basis of historical data is there for every group and phase of construction calculated an average delay. After that, the most relevant groups are selected. This reduced the total number of different groups which are included in the model to 33 which represent about 95 p.c. of all the delays. In the remaining groups are raw data inserted. As a last aid variable is historic information used (use data back to 2012) about how long time, it takes from the date of permission to the date of started for a building case. The criteria for success are still that the 1st estimate of a month must range inside +/- 5 p.c. of the "final" estimate for the same month (18th estimate).

Compared to the model from 2017 are there 3 major changes: 1. The actual model is a so-called generalized linear model of quasipoisson-family (previous was used linearly regression analyses). The model is trained from new for every running and will therefore be wiser and wiser as time goes and more data will be available. 2. The extra variable, "trend", is removed and replaced with interactions and a non-lineare expansion of the time variable through ”natural splines” 3. The calendar month enter again into as a factor, because analyses shows a clear variation in the delay pattern through the calendar year.

Small buildings: The government pass into law that small buildings (typically garages, carports and outhouses) until 50 m2 not - under certain conditions - longer are obliged to report construction (new construction, extension/rebuilding or demolishing) to the municipality. The event shall alone recorded in master data, in other words there will be no building project. In the construction statistic will these cases be picked up in master data and imposes a "building project" which handled similar other building project.

Adjustment

Seasonal adjustment is performed for construction permitted, construction started, ongoing construction and completed construction back to 1998 for the main groups of use of building (total floor space) and type of dwellings (number of dwellings).

Previously seasonal adjustment was only carried out for the four main time series: started construction and completed construction measured in floor area and dwellings, which were seasonally adjusted. In connection with publishing the data from 2th quarter 2012, a revised and extended seasonally adjustment was performed. Now all series are seasonally adjusted.