Scottish Employer Perspectives Survey 2021: Technical Report

Using the survey for analysis

Reporting

The headline results from the survey are available in an Official Statistics Publication on the Scottish Government website. Background data accompanying the Official Statistics Publication can also be found on the Scottish Government website. Accompanying 'Background Tables' include underlying data used in the report.

As this is the first EPS in the series to be undertaken since the start of the COVID-19 pandemic, comparisons with previous waves should be done with this in mind.

Sampling error and statistical confidence

Sampling errors for the survey results overall and for key sub-groups are presented in Table 10. Figures have been based on a survey result of 50% (the 'worst' case in terms of statistical reliability) and have used a 95% confidence level. Where the table indicates that a survey result based on all respondents has a sampling error of ±3.09%, this should be interpreted as follows: 'for a question asked of all respondents where the survey result is 50%, we are 95% confident that the true figure lies within the range 46.91% to 53.09%'.

As a note, the calculation of sampling error has taken into account the finite population correction factor to account for cases where we are measuring a significant portion of the population universe (i.e. even if two sample sizes are the same, the sampling error will be lower if in one case a far higher proportion of the population was covered).

These confidence intervals are based on the assumptions of random probability sampling and a normal distribution of responses.

Note: due to overlap between some ROAs, the sum of the number of interviews across the ROAs will exceed the total number of interviews. Population data is taken from the 2021 Inter-Departmental Business Register (IDBR) ‒ the latest available business population statistics published by ONS at the time of reporting.

Explaining variation between sub-groups in the report

Table 10shows the total number of interviews achieved across the entire sample. However, a number of measures are based only on a subsection of the sample. For example, questions on the provision of training were only asked of establishments that had provided training for their staff in the last 12 months (817 unweighted establishments). Table 11 shows common subsample bases used in reporting and their respective base sizes, both at an overall level and split by sizeband. Lower sample sizes increase the maximum sampling error, as shown in Table 10.