Appendix 5 - Survey methodology
Sampling and data collection
Using the June 2016 Agricultural Census (9) , a sample was drawn representing arable cultivation in Scotland. The country was divided into 11 land-use regions (Figure 14). Each sample was stratified by these land-use regions and according to holding size. The holding size groups were based on the total area of arable crops grown. The sampling fractions used within both regions and size groups were based on the areas of relevant crops grown rather than number of holdings, so that smaller holdings would not dominate the sample.
The survey covered rodenticide use during the 12 month period January to December 2016. Following an introductory letter and telephone call, data were collected by personal interview, telephone interview, email or post. When rodenticides were applied by a pest control professional ( PCP) the data were obtained from either the farm rodent control record book or by post/telephone interview from the contractor. If it was recorded that rodenticides were used but product data were not obtainable this was recorded as unspecified rodenticide use.
In total, information was collected from 240 holdings ( Table 6). These 240 holdings represent three per cent of the total arable holdings in Scotland ( Table 7). The data collected were; who conducted the baiting, product(s) used, bait type, weight applied, target and season of use. Information about use of non-chemical rodent control methods was also recorded.
All farmers and PCPs encountered in the survey were also asked to respond to a simple questionnaire containing questions relating to whether they had received training in use of rodenticides, their self-reported compliance with best use practice for rodenticides and their knowledge of rodenticide stewardship. Farmers were also asked to provide operation details about their farm, such as whether they kept livestock or had a grain store, to allow comparison in farm operation in relation to whether they used rodenticides.
It should be noted that, in relation to all data collected, responses are as reported by the rodenticide users and no attempt has been made to check their accuracy
Figure 14 Land use regions of Scotland (10)
Estimation of national rodenticide use
The figures presented in this report are produced by surveying a sample of holdings rather than conducting a census of all the holdings in Scotland. Therefore the figures are estimates of total rodenticide use for Scotland and should not be interpreted as exact.
National rodenticide use (holdings using rodenticides, rodenticide occurrence and weight) was estimated from the sample data by ratio raising. This is a standard statistical technique for producing estimates from a sample. This method involves multiplying the sample data by a factor dependent on the number of farms within each region and size group to match the data recorded in the relevant June Agricultural Census for arable crops. Due to small sample sizes the data from some regions were merged and a secondary adjustment factor was applied to the raising factors to account for region and size groups for which no holdings were sampled. Details of regions, size groups, raising and adjustment factors are presented in Table 8.
The remainder of the data (rodenticide target, use of non-chemical control methods, details of farm operation, compliance with best practice and knowledge of rodenticide stewardship) are unraised and represent the information collected from the sample.
Changes from previous years
In previous reports in this series, data about the type of rodenticide user ( i.e. farmer or PCP) were based on the proportions encountered in the sample surveyed. Whilst this was made clear in the reports, it was in contrast to the weight and occurrence data presented, which were estimates of use in the total population of arable farms. Due to improvements in our database, population estimates of user type are now available and replace the sample data in this report for both 2016 and historical data. This has resulted in differences in 2012 and 2014 user data presented in this report and in the original reports.
For the first time in this series of surveys we asked questions about knowledge of rodenticide stewardship, uptake of stewardship affiliated training and how rodenticides had been purchased post-stewardship.
As estimates are based on a random stratified sample of farms in each survey year and individual farms may be sampled more than once in the time series, there is no simple method of statistical comparison for estimated rodenticide use on arable farms over time. However, the percentage of farms using rodenticides, the percentage of farms on which baiting was conducted by PCPs and the percentage occurrence of first and second generation compounds have been analysed using Pearson's chi-squared test. The percentage occurrence of first and second generation compounds was analysed using the number of holdings as a base. These conservative analyses do not take into account the stratification, finite population sampling or common farms between years and are therefore less likely to find significant differences. All significant differences are highlighted in the text and tables of this report.
Data quality assurance
The dataset undergoes several validation processes as follows; (i) checking for any obvious errors upon data receipt (ii) checking and identifying inconsistencies with use and pesticide approval conditions once entered into the database (iii) 100 per cent checking of data held in the database against the raw data. Where inconsistencies are found these are checked against the records and with the farmer if necessary. Additional quality assurance is provided by sending reports for independent review. In addition, the Scottish pesticide survey unit is accredited to ISO 9001:2015. All survey related processes are documented in Standard Operating Procedures ( SOPs) and output is audited against these SOPs by internal auditors annually and by external auditors every three years.
Main sources of bias
These surveys may be subject to measurement bias as they are reliant on respondents recording data accurately. As surveys are not compulsory they may also be subject to non-response bias, as some farmers and PCPs may be more likely to respond than others. However, the use of a random stratified sample is an appropriate survey methodology and reserve lists of farms are held for each stratum to allow non-responding farms to be replaced with similar holdings.
Experience indicates that stratified random sampling, including reserves, coupled with personal interview technique, delivers the highest quality data and minimises non-response bias.