The SBR and the Business Statistics in Denmark

Sven Egmose  
Statistics Denmark

Introduction

Statistics Denmark is the key producer of statistics in Denmark. Statistics Denmark consists of four departments one for general internal services and three statistical departments defined along the subjects: Social statistics, business statistics and economic statistics. Most of the business statistics is naturally produced within the business department but also within the two other statistical departments some statistics based on information from business units are made up.

The departments are broken down in divisions more or less purely according to subjects. The business department is thus broken down in divisions according to industries with a remarkable exemption. One of the divisions contains both the Statistical Business Register (SBR) and the structural business statistics for all industries.

The argument for the division on SBR and structural statistics is the synergy effects between the work on maintaining the SBR and the in-depth work on making up the structural statistics. This has shown up to hold good. But also the short term statistics are intensive users of the SBR and as all the large and important business units also report to the short term statistics, a lot of updating information are gathered by the short term statistics and handed over to the SBR.

The SBR

The Danish SBR is very closely connected to the Danish administrative Business Register called CVR. The CVR contains basic information on all business entities in Denmark but only as regards the two unit types: Legal units and the LKAUs belonging to the legal units. The SBR is on line connected to the CVR, and the SBR contains a perfect copy of the CVR. But besides that the SBR comprises more unit types and more information for each of the unit types than the basic information in the CVR.

The CVR is updated by the House of Commerce and the Tax and Customs Authorities regarding the legal units within the private sector, and by Statistics Denmark regarding the legal units within the public sector. As to the information on activity code Statistics Denmark has a special right and duty to update this feature also for legal units within the private sector for the business entities surveyed by Statistics Denmark.

The LKAUs belonging to the legal units are manually (on line) updated by Statistics Danmarks and by the Factories Inspectorate. For Statistics Denmark the updating is done ad hoc when we get information on these units through our different surveys, and besides that Statistics Denmark once a year perform systematic checks on LKAUs belonging to legal units with more than one LKAU.

The Factories Inspectorate updates the LKAUs belonging to legal units on ad hoc basis, if they during their inspections at LKAUs get information that some of the data on a unit in the CVR is wrong, or a unit is missing in the CVR.

But most of the updating of LKAUs belonging to legal units is done automatic by the CVR system in connection with updating or start or closing of legal units.

Whenever a new legal unit is entered into the CVR system, the system attempts to generate a new LKAU belonging to the legal unit, but it is automatically checked within the system if this LKAU is a real new unit. If the system accepts it as a real new, a new entry of a LKAU belonging to the legal unit is fed into the CVR. If not, the system immediately corrects the new entry to be a takeover of an already existing unit.

Corresponding to the entering of new legal units also updating of the legal units are transferred to the LKAUs. Also updating of LKAUs are checked by the system to determine if it is really an updating of an existing LKAU or the reality is a closing of one LKAU and the start of another LKAU.

A part of these decisions taken by the system according to some built in rules are examined by the staff of the CVR and the SBR, and in a number of cases the staff overrules the decisions taken by the system.

This updating procedure for CVR and the on line connection between CVR and SBR ensures that Statistics Denmark always has access to information on all legal units  registered in the administrative systems and to the LKAUs belonging to these legal units.

Real time dating

In the CVR real time dating is attached to the units, the features of the units and the relations between units. In the SBR this is also the case, and not only for the information which is shared with the CVR. Also for the statistical unit types, the features of these units and the relations, real time dating exists in the SBR.

This is of course a considerable complication of the system compared to a system with only information regarding the current situation. But it also opens some significant possibilities in using the SBR for the statistics both regarding the use for statistics concerning a given period or a given point of time and regarding the use of SBR as a basis for demography statistics.

The real time dating within the SBR opens the possibility that you at any time can make an extract from the SBR with the best possible information regarding a specific time in the past. If you during the summer 2004 want an extract of all the enterprises which were active at the end of year 2003 you can e.g. July 1, 2004 make an extract from the SBR with the best possible information regarding the state of the art at December 31, 2003.  If you come again one month later you can August 1, 2004 get an extract from the SBR with even better information regarding the state of the art at December 31, 2003, because some corrections which in reality were valid before December 31, 2003 has been fed into the SBR during the month of July 2004 with real time dating stating from what point of time the corrections in reality were valid.

This possibility offers opportunities for always using the best quality information available for the statistics. But at the same time it constitutes a challenge towards the coherence of the statistics, and coherence between statistics is a very important quality component. For the intensive users of statistics this is maybe even the most important quality component.

The use of the SBR

Within Statistics Denmark virtually all the business statistics are based on information from the SBR. For statistics produced by using data from administrative sources the population and the features (activity code etc.) of the business units are drawn from the SBR. For statistics relying on surveys the same goes regarding population and features but besides that also the sample for the survey is drawn from the SBR.

In Statistics Denmark the SBR is placed at the disposal for the different statistics, and the way to use the SBR is to a rather high degree left to the different statistical branches. Whenever a statistical branch plans to make a statistics this branch orders an extract from the SBR, and after some consultations and clarifying questions the staff of the SBR draw an extract from the SBR and place it at the disposal for the statistical branch.

Quality of specific statistics versus coherence

The statistician in charge of a given statistics wants to make this specific statistics with as good a quality as possible. And with a SBR containing real dating of the information, the best quality data regarding a given point of time is the latest information. The data on activity codes for enterprises in the SBR by e.g. the end of 2003 will be better tomorrow than it is today, because some corrections might be made in the SBR. Therefore the statisticians want the withdrawal from the SBR for their statistics to be made as late as possible. And as the different statistics regarding a given period are made up at different points of time, the SBR data used for the different statistics will be different. This means that the statistics as such are as good quality statistics as possible. But this is ensured at the cost of coherence between the statistics.

Coherence is an important goal for the statistics to aspire at, but it is a hard job to convince the single statistician to use the measures necessary, i.e. that some of the statistics will as such have a lower quality than it could have.

Coherence is a strategic goal

One of the strategic goals for Statistics Denmark is to produce coherent statistics. For the business statistics the decision is to go for coherent statistics by using the same unit types as statistical units in different statistics and by using the same versions of the SBR as basis for the grossing up population and the background variables for different statistics. Nevertheless, the standardisation will not be driven that far as to use only one statistical unit and one version of the SBR for all statistics.

Unit types

The different statistical units are suitable for different statistics. What is the intention is to use LKAU as statistical unit for some statistics (the KAU and the local unit are aggregations of LKAUs), and to use enterprises for other statistics. For the time being some statistics in fact use other unit types such as legal units or even administrative (taxation) units for the statistics, but it is the strategy of Statistics Denmark to phase out these unit types and to use only the two statistical units (including the aggregates mentioned) for the statistics in the near future. The standardisation of the unit type used for statistics is of course a very important tool for establishing coherent business statistics.

Frozen versions of the SBR

The standardisation of the versions of the SBR is almost as important. But as mentioned above the possibility to improve the background variables for the statistics by waiting to draw extracts from the SBR until the latest moment possible for the single statistics ¨C and sometimes even making corrections for the population of this statistics by the branch statisticians themselves ¨C is an opportunity which the branch statistician will not let go without striking a blow.

To move towards more coherent statistics the specific interests of the branch statistics must be overruled, and the top management of Statistics Denmark has decided that this shall be done. The starting point is some of the essential yearly structural business statistics, where we are aiming at using the same frozen version of the SBR as basis for the statistics. Regarding the short term statistics the problem of introducing frozen versions of the SBR is harder than regarding structural statistics. The reason for this is that more corrections come into the SBR regarding the nearer past than regarding the farther past. For the yearly structural business statistics the withdrawal from the SBR will not have to be drawn before 10 months or so after the year in question. At that point of time corrections regarding the relevant period is thinning out. For some short term statistics it will be necessary to make the withdrawal from the SBR already within the first month after the period in question. Other short term statistics will be able to wait maybe two months to draw the background variables from the SBR. During that period rather many corrections regarding the relevant period are fed into the SBR.

To implement the use of the same version, a frozen version, of the SBR for the statistics requires a real good quality of the SBR. Through the process of producing the different statistics also the population and the background variables from the SBR are gone through by the branch statisticians. If the branch statisticians run into cases of erroneous background variables, such as activity codes which ought to be corrected, the pressure grows to give up the frozen version of SBR. Up till now such pressure has caused that Statistics Denmark has not succeeded in using frozen versions of the SBR for the statistics even if the intention has been to do so.

The quality of the SBR has not been good enough for sticking to frozen versions of the SBR for the statistics. The subject matter statisticians have been able to point at too many erroneous data, e.g. on activity code, in the SBR for sticking to a frozen version of the SBR. Now a hard work on improving the quality of the SBR data is taken on, and that will pave the way for using the same version of the SBR for the statistics. This will be done step by step, and in the first hand the most important structural business statistics, the statistics on employment and the statistics on accounts, will be based on the same version of the SBR

This will of course improve the consistency between the two statistics. Another important result is that it will improve the possibilities for integration on micro level between the two statistics. Such integration has been done also in former years, both for the sake of Statistics Denmark¡¯s integrated statistics on enterprises, and for the sake of researchers outside statistics Denmark who have access to such data for scientific analyses. This data integration has been very troublesome, and one of the main reasons for that is that different versions of the SBR have been used for the two statistics.  The expectation is that the use of frozen version of the SBR can be spread out to all structural business statistics within the next few years.

Regarding the short term statistics the level of ambition is also to go for frozen versions of the SBR for the statistics. It will naturally not be the same frozen version as the version for structural statistics because of the difference of time. But the acknowledgment is growing regarding the importance of the versions of the SBR used for the statistics. An internal working group within Statistics Denmark has this autumn released a report on the continuation of the work on quality of the short term statistics. One of the main points raised by the working group is to go for the same version of the SBR for the different statistics. This means that focus now will come on defining which version of SBR to use for the different short term statistics aiming at one frozen version for all short term statistics regarding a given period of time.

Corrections to frozen versions

When introducing the use of frozen versions of the SBR for the statistics the intention is not to freeze the versions completely. If severe faults in the SBR data are detected through the process of making up the statistics, corrections to the ¡°frozen¡± version can be accepted, but the corrections must be approved by a high level group within Statistics Denmark to be allowed. This high level group shall also determine if corrections shall cause that already published statistics shall be published in a revised version.

Samples for questionnaires

Regarding drawing of samples for surveys the need for standardisation for reasons of coherent statistics is not that pressing. As long as the population for grossing up and the background variables used in the statistics are standardised by using the same version of the SBR, it is not very important if the samples are coordinated. What is important is that the data for the samples are valid and correct, so that questionnaires can reach a sample of business units who can answer the questionnaires. The coordination must be produced through the grossing up procedures.

But coordination of survey samples might be necessary for other reasons such as efficiency and response burden matters.

Up till now the different statistical branches within Statistics Denmark have drawn their own samples from the SBR. The ordinary way to draw samples from the Danish SBR is that the staff of the SBR extracts the total population e.g. all units within retail trade and that the branch statistician draw the samples from this population.

One advantage in using this procedure is that the statistician who is responsible for a statistics has a detailed knowledge on the sector which the statistics is about, and hence he or she is aware of what is happening or about to happen regarding the business units within this sector. Nevertheless, this way of drawing samples is probably not the most efficient way to do it. For that reason this way of organizing sample drawing has become under pressure. But another argument is even more convincing than the cost argument. The response burden has become a very high focused topic at least in Denmark, and in particular the burden on small enterprises is in focus.

The SBR is an important tool to minimize the response burden. A good quality SBR as the frame for statistics provides the opportunity to restrict the sample sizes. This can be done without special coordination of the samples. But if you in addition want to ensure a certain distribution of the burden, a coordination of the sample drawing is a necessity. The distribution of the burden has until  now not been dealt with in a systematic way, but it has become more and more in focus that apart from the wish to reduce the statistical response burden, the remaining burden should be distributed in a fair way. I.e. the small business units should not be bothered by a lot of questionnaires. To comply with this Statistics Denmark has committed itself to ensure that from next year the small enterprises will at most be ordered to answer questionnaires for up till 3 surveys a year.

To ensure this, coordination of the drawing of samples is necessary. The coordination can come along in different ways, and as it for the recent year only has been a matter of some few hundreds of small enterprises which had to fill in questionnaires for more than 3 statistics, the coordination can be taken care of manually. But even if it only concerns rather few enterprises the manual handling is not efficient. The efficient way to ensure the coordination is to centralize the drawing of samples, and this will no doubt be the future situation for business surveys performed by Statistics Denmark.