Cereal and oilseed rape harvest: Methodology

Accuracy and reliability

This section discusses how accurately and reliably these statistics portray reality.

June Agricultural Census Area Estimates

Data quality and assurance measures for the June Agricultural Census are available alongside the Scottish Agricultural Census results.

First Estimates

These data are intended to provide a prediction or intermediary result and are replaced by the release of final estimates. 

First estimates are based on data collected in September while the harvest is still underway in areas of Scotland. Given the unpredictability of the weather and the knock-on effects this has on production (e.g. poorer yields in later harvested crops which have been subjected to longer periods in unfavourable growing conditions), providing early estimates of cereal yields and production is a challenging task.

The collection of yield estimates prior to the completion of the harvest may lead to a level of inaccuracy with these estimates which is difficult to estimate because of this unpredictability.

Results will have bias towards farms that have completed their harvest earlier in the year and include estimates from farms that may not yet have completed. As the harvest continues it is inevitable that regional and national production and yield reports will become more reliable.

The Crop Report Meeting and discussion of results with the industry bodies that attend provides quality assurance and validation to the initial estimates that are gathered across Scotland.

The nature of the industry bodies involved in the production of provisional estimates also means that results are likely to be more representative of commercial cropping farms and less representative of farms growing crops for on-farm uses. This is likely to contribute to first estimates tending to overestimate production and yield compared to the final estimates.

Comparison of first and final estimates

This section compares first and final estimates of the harvest. The purpose of this section is to quantify the size and direction of the differences between the two estimates in order to give an indication of the robustness of first estimates.

The main reason for differences in the first and final production estimates are differences between first and final yield estimates. A table of the full breakdown of the differences between the estimates is provided in the Final Estimates Publication.

Figure 1 shows that in the last ten years the first estimate of the total cereal harvest has typically been within five per cent of the final estimate.

In 2018 the first estimates were between two and sixteen per cent higher than the final estimates. This is due to fewer farm results being available for the first estimate, especially for wheat.

Figure 1:  Cereal production, percentage comparison of first and final estimates, 2014 to 2023

In some years provisional June Agricultural Census area figures are used to calculate first production estimates. These area estimates have been slightly different from the final June Agricultural Census areas used to calculate the final production estimates. However, these differences have generally been small (less than five per cent) and are not a main contributor to differences in the production estimates.

Final estimates

Final estimates are collected when most farms will have completed their harvest.

The CPDS is a disproportionate stratified random sample of around 500 farms in Scotland, stratified by region. In 2023 the sample size was increased to around 1,200 farms. Oversampling in some groups is designed to ensure that the result has good representation of all groups.

To provide reliable estimates of the year-on-year changes in production, the CPDS is carried out at the same time each year. The reference date for the CPDS, the date at which respondents are asked for production information, is the 31st October each year.

Typically, at the end of October the vast majority of the Scottish cereal and oilseed rape harvest is complete, allowing for reliable estimates to be made, although respondents are also asked to make estimates for any crop still to be harvested.

Returns are validated to ensure there are no obvious errors, assess extreme values and cross checked against census area estimates to ensure totals match.

As a sample survey, response rate has an impact on accuracy and reliability. Response rates and sample sizes in recent years are shown in Table 2. Between 400 and 600 holdings in Scotland are sampled each year. Many holdings grow more than one crop, so returns by crop are higher than the number of returns by holding.

Table 2: Sample size and returns to the Cereal Production and Disposal Survey (CPDS)

The results from the CPDS have a margin of error associated with them, reflecting the error resulting from sampling. Sampling error is the difference between the estimate derived from a sample survey and the true value that would result if a census of the whole population were taken under the same conditions.

The sampling error can be estimated and used to produce confidence intervals around the survey results. These intervals tell us the range of values within which the true value lies, with a given degree of confidence. The intervals below are 95 per cent confidence intervals; this means that if the sample survey was repeated a large number of times, 95 per cent of the resulting estimates would lie within the intervals around our sample estimates.

In 2023, there is a 95 per cent chance that the true production value for all cereals in Scotland will lie within the range of 3,070 thousand tonnes ±126 thousand tonnes. Table 3 gives the 95 per cent confidence intervals for the 2021 CPDS.

Table 3: 95 per cent confidence intervals for the 2023 Cereal Production and Disposal Survey (CPDS)