The Scottish Government

« Previous | Contents | Next »

Listen

3.0 Comparison of SAF and JAC holding/business populations

3.1 Introduction

The objective of this section is to assess the adequacy of coverage that SAF would provide in terms of numbers and size of holdings, farm types and geographic distribution. The adequacy of the SAF dataset in terms of its ability to generate answers to the specific questions on areas and numbers within the JAC is addressed in sections 3 and 4. Following a description the materials and methods used (Section 0) the chapter presents analysis comparing how many holdings/businesses are present only in SAF or JAC and which are common to both (Section 0) - these are presented as national summaries and broken down by Farm Type, NUTS4 region and Farm Size represented by Main and Minor holdings as defined by the June Census.

3.2 Materials and Methods

Datasets

The SAF data used for this analysis is a "mirrored" copy of selected tables from the SGSIACS database held at the Macaulay Institute on an Oracle relational database. The database holds complete SAF records for 2005, 2006 and 2007. The latter had been a partial dataset when originally supplied by RPID and was updated with the most recently available data on January 31st. The JAC data was supplied by Analytical Services as "flat file" tabulated data, with one row per holding and a column for each JAC item. The JAC dataset contains both returned and imputed data. Where imputed data is included then the nomenclature JAC(All) is used and where only returns are used then JAC(M&m). Tables to hold the JAC data (one table per year) were added to the MI Oracle database. This permitted the combined used of the two datasets - for example the use of Farm Types defined in the JAC datasets to be used to group sets of holdings within the SAF.

Integration Issues

The distinction between holdings and businesses is a key issue in constraining the comparisons that can be made between JAC questions and SAF items. As noted in Section 1, for the land use related JAC questions information is returned on a field-by-field basis by the SAF Field Data Sheets ( FDS) ³ - see Figure 1. There are also known field-to-holding relationships held in the SIACS areas_of_land table ⁴ that allow aggregation of the SAF ( FDS) data to holding level - see Figure 2. Direct holding to holding comparisons can therefore be made for these items between SAF and JAC. The holding level results can also be grouped by Farm Type and by NUTS4 region. The SAF data used for these comparisons is identified as SAF ( FDS).

Figure 1. Schematic diagram of JAC and SAF data comparion

For JAC livestock questions the situation is more complex. JAC records livestock numbers at the holding level, but SAF returns for livestock numbers need only be made at the business level which can comprise multiple holdings. Figure 2 illustrates this situation using the Macaulay Institute's research stations as an example - all livestock for three holdings are reported through the Glensaugh holding in Aberdeenshire. This means that for comparisons to be made between the SAF and JAC for livestock items, the JAC data need to be grouped into businesses. When business level JAC data is referred to the nomenclature JAC(All-b) ⁵is used. It is possible to link approximately 33,000 JAC holdings to businesses (using relationships stored within the SIACSCDM⁶). Table 1 reports the number of JAC holdings that can be resolved to businesses (33,284 holdings (65%) are resolved to 28,365 businesses in 2007). The inability to resolve all holdings to businesses (18080 remain "orphaned" in 2007 ⁷) may weaken the utility of the SAF as an alternative or even complementary data source for livestock questions.

Table 1 - Holdings Resolved to Businesses

Even where it is possible to link holdings to a business there remain three of outstanding issues for the comparison or SAF(L) and JAC(All-b).

1. It is not certain if all the holdings that make up a business (with a BRN) must be identified in the SIACSCDM. Not including all holdings in a business could have seriously distorting effect on aggregation processes and thus on the comparisons between SAF(L) and JAC(All-b)
2. Since some businesses are be made up of holdings with multiple different farm types it is not possible with the data currently available to assign them a single farm type on a consistent basis.
3. Since businesses may be made up of holdings widely separated in geographic space it is not possible to assign them to a single NUTS4 region.

The consequence of the first issue is a further weakening of the comparability of the two datasets and the remaining three issues mean that is that it is not possible to provide a meaningful comparison of livestock figures from JAC(All-b) and SAF(L) by size, Farm Type or at NUTS4 regions (See Sections 3/ 4).

Figure 2. Schematic Diagram of Spatial Identifiers

Analyses and Outputs

The outputs from the analyses are stored as views within the ORACLE database and can be exported as flat text files for use within spreadsheets or as tables in documents. The illustrations in the document other than maps are generated in MS-Excel. The maps in this document are implemented in ArcInfo 9.2 and are used to illustrate the geographic coverage provided by both datasets. The NUTS2 region base maps were provided for research use by Edina ⁸. On these maps the shading is used to allow quick comparisons between regions. An exact value of the attribute being mapped is added as a symbol in the centre of each region. The smaller NUTS4 regions in Scotland's "central belt" are shown at a larger scale.

3.3 Results

For the summaries and case-studies the comparison is made between JAC (All) and SAF( FDS) or JAC(All-b) and SAF(L). As noted previously JAC (All) includes imputed data while the SAF based data sets do not. The JAC imputation uses information provided by a holding in a previous year and rolling forward this information using average trends in a particular strata, based on farm size, farm type and geographical location. This provides an imputed value which is partially informed by previous actual data. The imputed data is included in the analysis since Analytical Services wish to asses the implications of substituting SAF returns data (without imputation) for JAC data as it exists at present (i.e. with imputed values). The effectiveness of using JAC imputation procedures with SAF data to create a "complete" coverage lies beyond the scope of this research.

It is possible to identify those holdings for which SAF is returned but JAC is imputed. Such holdings would represent potential additional data that could be made available to JAC. Comparison between holdings where there are returns, termed JAC(M&m), and SAF( FDS) and SAF(L) could bring out where the JAC and SAF datasets overlap, where JAC data alone is available, where SAF data alone is available and where only imputed JAC data is available. Where appropriate supplementary JAC(M&m) figures are included in the summary data to assist with interpretation.

Comparisons of JAC(All) with SAF( FDS)

Table 2 presents a national level summary comparing holding numbers in JAC(All) and SAF( FDS). The table also presents the percentage of holdings covered by SAF( FDS). Table 3 presents the same comparison but on an area basis. For 2007 the coverage of JAC(All) by SAF( FDS) was 59% of holdings and 87 % of area. Again for 2007 coverage of major holdings was larger than for minor holdings at 81% versus 36% of holdings and 89% versus 61%. This reflects the fact that SAF is returned more often by larger units. It is worth noting that in terms of holdings and area the coverage of JAC(All) by SAF( FDS) is improving. RPID staff have also indicated that coverage is likely to increase in 2008 since a wider range of holdings will submit SAF as part of making applications for LMCs.

Table 2 - Count of Holdings and % Coverage JAC(All) by SAF( FDS)

Table 3 - Area of Holdings and % Coverage JAC(All) by SAF( FDS)

As a case study the results for 2007 are presented in more detail. The breakdown in the overlap of the holding populations between SAF( FDS) and JAC(All) is shown in Figure 3. Note for these figures the percentages are of all holdings (i.e. JAC(All) plus those in SAF( FDS) only). There is an apparently substantial mismatch in populations (43% are not common). Of that 43%, however, over half are returns from minor holdings. This means that in terms of area rather than holdings the overlap is substantial (82%) with most of that overlap being in main holdings (79%) see Figure 4. There is also some potential to capture more information on land use from the SAF( FDS) only (6% of area). The 31% of the population that are minor holdings in JAC but do not appear in the SAF equate to only 2% of area, but may still be significant for particular activities.

Overlap in Holding Populations - JAC (All) and SAF (FDS)

Figure 3.

Overlap in Holding Area - JAC (All) and SAF (FDS)

Figure 4.

Comparison of JAC(All) or JAC(All-b) with SAF (L)

The SAF(L) businesses are those identified as claiming under the SFPS. Comparing their coverage with the JAC is complicated by the fact that the livestock numbers for SAF are collected at a business level. This means that some comparisons are only possible between SAF(L) and JAC(All-b) (those JAC holdings that can be resolved to businesses). For example Table 4 presents the count of businesses for the two datasets and the percentage coverage of JAC(All-b) businesses by SAF(L). The coverage of 69% could be interpreted as superior to the 59% overall coverage of JAC(All) by SAF( FDS) but inferior to the 81% of coverage for JAC(All)-main by SAF( FDS). Comparisons on this basis are potentially misleading, however, since any percentage of JAC(All-b) is a fraction of a significantly smaller whole since for example JAC(All-b) covers only 65% of the JAC(All) holdings (Table 1) and 89% of the area of JAC(All) for 2007 (see Table 5). The comparisons of SAF(L) with JAC(All-b) using counts of businesses are retained here, however, since they inform the interpretation of the individual business-by-business comparisons of livestock numbers presented in Section 3. A more reliable comparison of coverage is provided by using the areas of land covered by each of JAC(All), JAC(All-b) and SAF(L) - see Table 6. From this table it can be seen that SAF(L) provides a comparable coverage (87% of area) for JAC(All-b) (business to business) to that provided by SAF( FDS) for JAC(All) (holding to holding). The coverage of JAC(All) by SAF(L) is, however, significantly inferior, at 78%. Note also the area of land in JAC(All) that cannot be resolved to any business.

Table 4 - Count of businesses in JAC(All-b) and SAF(L)

Table 5 - Areas of JAC(All) and JAC(All-b)

Table 6 - Area of businesses in JAC(All), JAC(All-b) and SAF(L)

The breakdown in the overlap of the business populations between JAC(All-b) and SAF(L) for 2007 is shown in Figure 5. Note for this figure the percentages in the chart annotation are of JAC(All-b) plus those in SAF(L) only, and thus do not exactly match the percentage figures used as summaries in Table 4. The overlap in the populations of businesses for JAC(All-b) and SAF(L) is 66% while the equivalent coverage of JAC(All) holdings by SAF( FDS) is 57%. As noted for the summaries above this superior coverage is of a smaller population and needs to be interpreted with care. When the comparison is done on the basis of area (see Figure 5), and the area of land within JAC(All) that cannot be resolved to businesses is included (10.8%) then the coverage of JAC(All) by SAF(L) is at 76% inferior to the 82% for SAF( FDS). Without including the areas of the unresolved JAC(All) holdings the coverage of JAC(All-b) is 85%. While there is some potential for additional information from SAF(L) - 4% of businesses and 3% of area there are a substantial number of businesses that appear only in the JAC - 30% of businesses and 11% of area. To repeat there is the need for great care in interpreting both the population and area-based comparisons of SAF(L) and JAC(All) since there are substantial numbers of holdings that cannot be included in the analysis ⁹ (for the comparison of business numbers) and there are differences in the inclusions and exclusions of areas defined for SAF and JAC.

Overlap in Business Populations - JAC (All-b) and SAF (L)

Figure 5.

Overlap in Business Area - JAC(All), JAC(All-b) and SAF (L)

Figure 6.

Farm Type Analysis - JAC(All) vs. SAF( FDS)

The farm type analysis is presented only for the SAF( FDS) for the reasons outlined in Section 1.2.2 (Integration Issues). Tables 6 to 8 show a summary of the coverage per Farm Type for the years 2005-7. Note that since it is not possible, within the scope of this project, to assign farm-types to those holdings which appear only in SAF( FDS) then it is only possible to compare JAC(All) with the SAF( FDS) holdings common to both datasets - the Common (All) column in the tables below.

Table 7 - Farm Type Analysis 2007

Table 8 - Farm Type Analysis 2006

Table 9 - Farm Type Analysis 2005

Farm Types Analysis - 2007 Case Study

Figures 7 and 8 provide a further break down of the per Farm Type coverage of JAC(All) provided by SAF ( FDS). Figure 7 presents the breakdown as the count of holdings in both datasets and those only within JAC(All). Figure 12 presents the percentage of coverage for each Farm Type.

Coverage of JAC(All) by SAF(FDS)

Figure 7

SAF (FDS) % of JAC(All)

Figure 8.

The differences in the proportion of coverage of JAC(All) by SAF( FDS) reflect the activities administered through SAF - with strong coverage of LFA farm, cereals, dairy , general cropping and mixed farming. Coverage of horticulture and specialist livestock are lower since these have not been recipients of support (?) and the low coverage of Other perhaps reflects the fact that many of these holdings are small holdings who do not make SAF applications.

Imputation

An average of 25% of JAC(All) holdings contain only imputed data (this percentage does not include minor holdings unless they have not made a return for more than three years). Figure 12 shows the breakdown by Farm Type of the percentages imputed, with a range from 18 % for Specialist Poultry to 36% for Dairy.

Overall Imputation % of JAC(All)

Figure 9

This means that there is some potential for SAF( FDS) data to add value to the JAC by providing values where otherwise only imputed values are available. The relative importance of these sources is presented in Figure 13. This presents as a percentage of JAC(All) holding numbers the following:

1. Holdings with returns from both SAF( FDS) and JAC(All) - i.e. the overlap of JAC(M&m) and SAF( FSD), termed Common;
2. Holdings with SAF ( FDS) returns but only imputed JAC data;
3. Holdings with JAC(M&m) data and no SAF( FDS) returns and
4. Holdings with imputed JAC data only, JAC(I).

Coverage of Each Dataset

Figure 10.

From this figure it is possible to identify cases such as Cattle and Sheep ( LFA), Cereals, Dairy and Mixed Farming where use of SAF( FDS) could lead to significant reductions in the need for imputation, no Farm Type with more than 20%.

NUTS4 Regional Analysis - SAF( FDS) vs JAC(All)

Tables 9-11 present for the years 2005-7 the break down by NUTS4 region of the coverage of JAC(All) provided by SAF( FDS).

Table 10 - NUTS4 Breakdown 2007

Table 11 - NUTS4 Breakdown 2006

Table 12 - NUTS4 Breakdown 2005

NUTS4 Regional Analysis - 2007 Case Study

The NUTS4 regional analysis comparing the JAC (All) and SAF ( FDS) for 2007is presented as a series of maps in Annex III to this document.

« Previous | Contents | Next »