Creating a farm register the population

When a statistical register is created, all relevant sources should be used so that the coverage will be as good as possible. When microdata from different sources are integrated many quality issues become clear that otherwise would have escaped notice. However, a common way of working is to use only one administrative source at a time. Figure 2.5 shows the consequences of this. The Swedish Business Register has been based on only one source, the Business Register of the Swedish Tax Board. Between 1996 and 1997 the rules for VAT registration were changed. Earlier enterprises with a turnover smaller than 200000 SEK were not registered for VAT, but from 1997 these are included in the Tax Board's register. Quality improvements can explain the change between 2005 and 2006; those who work with the Business Register now also use typology data from the Farm Register. The definition of NACE 01 in the Business Register contains more activities than the definition used for the population in the Farm Register, which can explain the difference during 2007; if detailed NACE is used this

Figure 2.5 Number of agricultural enterprises in Sweden according to the Business Register (BR) and the Farm Register (FR).

Figure 2.5 Number of agricultural enterprises in Sweden according to the Business Register (BR) and the Farm Register (FR).

difference can be reduced. But it is clear that the Business Register series is of low quality - the Business Register has both undercoverage and overcoverage, the quality of NACE for agricultural enterprises has been low and administrative changes disturb the time series pattern.

The errors in the Farm Register series are not so dramatic, but still they can be seen in Figure 2.5. The traditional Farm Register was replaced by a register mainly based on IACS during 2003. Due to undercoverage (not all farmers could apply for subsidies) the value for 2003 may be too low. During 2005 the system for subsides was changed and the undercoverage was reduced. The IACS register may have undercoverage and here, too, we see that changes in the IACS system will disturb the time series. There are two kinds of administrative systems - some systems concern all (e.g. everyone has to report and pay tax), while other systems concern only those who apply or use the benefit or system (e.g. only those who benefit from subsidies are found in the IACS register). This means that different administrative registers can be more or less sensitive to undercoverage.

To be able to create a Farm Register harmonized with the Business Register the following problems must be solved:

1. A Farm Register that is not disturbed by undercoverage and changes in the administrative system should be created. This is discussed below.

2. The under- and overcoverage of agricultural enterprises in the Business Register that we have today must be eliminated. This includes the harmonization of NACE in the Business Register with the typology in the Farm Register. This is discussed below.

3. The administrative units in taxation data, the administrative units in IACS applications and the statistical units in the Farm Register and Business Register must be linked so that different sources can be integrated. This is discussed in Section 2.5.

4. All enterprises with some agricultural production should be included in the register population of the Farm Register with a link to the Business Register, and the importance of agriculture should be incorporated in the registers. The aim should be to have the importance of all NACE activities measured in percentage terms for each enterprise and establishment. This is discussed in Section 2.5.

We will illustrate these issues with examples from Statistics Sweden. We think that similar problems also exist in most other countries and the methods to solve them will be the same.

We begin with a discussion of the farm register design. Today the Swedish Farm Register is of high quality, updated yearly with IACS data and every second or third year with the Farm Structure Survey that contains questions aimed at improving register quality with regard to the statistical units. However, Business Register data are not used to update and improve the Farm Register and IACS data are not used to update the population in the Business Register. This means that available data are not used in an optimal way today and also that these two registers are not coordinated and that statistics based on them are not as consistent and coherent as they could be. The quality of the Farm Register is today mainly dependent on the administrative IACS system with regard to coverage. This is a potential risk - in the future the quality of this administrative source may change.

Within the entire statistical system, the main task of the Business Register is to keep track of the population of all kinds of enterprises and organizations. So the population of agricultural enterprises should be defined by the Business Register and all sources available to do this should be used by the Business Register. An important source is the weekly updates of the Tax Board's Business Register that are sent to Statistics Sweden's Business Register that are not used for the Farm Register today.

From a theoretical point of view, the Business Register should also be based on the sources used for the Farm Register, and the Farm Register population should also be based on the sources used for the Business Register. Then data would be used in the most efficient way, we would get the best possible quality and agricultural statistics would be consistent and coherent with other kinds of economic statistics.

Since 2006, we have created a calendar year version of the Business Register that consists of all enterprises and organizations active for some part of a specific calendar year. For this register we use all available sources and the register describing year t can be created during January of year t + 2. Activity is defined as non-zero values in at least one administrative source regarding year t. This calendar year register has very good coverage and we plan to use it as a standardized population for the economic statistics produced for the yearly National Accounts. We have also used the calendar year register to study coverage errors in our present economic statistics, and it is planned to use our findings to improve quality. This project is described in Wallgren and Wallgren (2008).

The IACS and the Farm Register should also be used when this calendar year register is created - the population and NACE for the agricultural part of the register will then be as good as possible. But this calendar year register cannot be used as frame for samples or censuses. This means that frames must be created almost as they are today, but with better sources. Figure 2.6 illustrates the production processes for 2007. The points in time when the frames for the Farm Structure Survey and the Structural Business Survey were created are shown above the time axis. Based on all administrative sources below the time axis, the calendar year register for 2007 was created during January 2009. With this calendar year register the preliminary estimates for the Farm Structure Survey and the Structural Business Survey can be revised. The revisions give information about the frame errors in preliminary estimates. This knowledge is the starting-point for the work to improve the Farm Register and the Business Register. We expect that the frames produced with these

Frame for Yearly FSS 2007 for SB


Calendar year register 2007 frame based on ALL sources S 2007 All enterprises active during 2007

NA 2007 data delivery

7 V i

1 1 1 1 1 1 1 1 1 1 1


A 2008 A A 2009

Deliveries of administrative data for 2007

Figure 2.6 Frames used for the Farm Structure Survey (FSS) and Structural Business Survey (SBS) and the corresponding calendar year register.

Table 2.2 Coverage errors in the Business Register (BR) compared with the Farm Register (FR) data for 2004 and 2005.

Number FR

of units in BR


Coverage errors in SBS

SEK millions


NACE 01 in BR

6 787



1110 1.8%

not active



NACE 01 in BR


Yes, not in 01

not in FR


In FR and BR



Yes, in 01


In FR not in BR

8 696

Yes, in 01


4425 6.8%


Total number: 75 808 138 39?

Total number: 75 808 138 39?

registers will gradually improve. As a result, more and more Business Register data will be used for the Farm Register and the dependence of IACS will become smaller.

We now turn to under- and overcoverage in the Business Register. Selander (2008) describes the undercoverage in the Business Register with regard to agricultural enterprises. Statistics Sweden (2007) describes the corresponding overcoverage. In Table 2.2 the population in the Farm Register has been matched against the units in the Business Register.

In the Farm Register there are 75 808 units and in the Business Register there are 138 397 units coded as active in NACE 01. Of these 138 397 units 6787 (line 1) were not found in any administrative source for the reference year, thus they represent overcoverage in the Business Register. In the Structural Business Survey (SBS) these inactive enterprises were treated as non-response and positive values were imputed. This gave rise to an overcoverage error of 1110 million SEK, or 1.8% of the total production value for NACE 01.

A further 64 498 units in the Business Register (line 2) were coded as active within NACE 01. However, as they did not belong to the Farm Register, we suspect that the NACE classifications in the Business Register of these enterprises in most cases are wrong - they are not active in NACE 01 but in another branch of industry instead. These enterprises thus represent overcoverage with regard to NACE 01 and undercoverage with regard to other unknown branches of industry.

A Further 8696 units in the Farm Register were not found in the Business Register (line 4). We can classify these as undercoverage in the Business Register, and due to this undercoverage the estimate of turnover in the Structural Business Survey was 4425 million SEK too low, or 6.8% of the total turnover value for NACE 01. We estimated this coverage error with information in the Value Added Tax Register during a project in which we analysed the quality of the Swedish Business Register. This project is reported in Statistics Sweden (2007) and in Wallgren and Wallgren (2008).

The conclusion is that the Swedish Business Register must be redesigned. More administrative sources, such as VAT, monthly payroll reports to the tax authorities and yearly income declarations from enterprises must be used to improve the coverage. The IACS and CDB registers should also be used to improve the agricultural part of the Business Register.

The typology in the Farm Register should be used to improve the NACE codes in the Business Register. The enterprises in the Business Register, today coded as NACE 01 but missing in the Farm Register, should get new NACE codes after a register-maintenance survey.

After these improvements to the Business Register it will be possible to integrate the Farm Register with the Business Register and use a large number of administrative sources for agricultural statistics in the same way as we today use many administrative registers to produce social and economic statistics.

Was this article helpful?

0 0

Post a comment