Official Statistics

Methodology for Coronavirus Job Retention Scheme statistics: secondary analysis of ended furloughs

Published 22 October 2020

1. Methodology

Please see the main statistical publications for Coronavirus Job Retention Scheme (CJRS) and Earnings and employment from Pay As You Earn Real Time Information (PAYE RTI) for detailed methodological information.

This release contains secondary analysis of the CJRS data and PAYE RTI data up to 31 August 2020 (first published on 18 September 2020 for CJRS and on 15 September 2020 for PAYE RTI). Results are therefore not updated, nor necessarily consistent, with more recent periods.

This analysis only includes employments in CJRS that could be matched to RTI to obtain their pay data for the month in which that employment was last furloughed. Employers making claims for 100 or more employees are required to submit the details of the employees furloughed in a spreadsheet-type file. While these claims have been processed from a customer service perspective, the processing of this information for these statistics has been complex and the processing of data on some employments has not been completed. Of 9,654,000 employments in the CJRS dataset, 9,532,000 matched to RTI (98.7%).

The CJRS data defines an employment as an individual in a PAYE scheme, whereas the RTI data can contain multiple employments for an individual within the same PAYE scheme. Where the CJRS data matched to more than one entry in the RTI data using National Insurance number and PAYE scheme reference number (0.8% of the matched data), we used the probabilistic imputations from RTI (see below) and combined them to determine the probability that at least one of those RTI employments remains on payroll.

CJRS claims submitted by employers include information on the dates of furlough periods. We define the last reported furlough month as that in which the most recent furlough end date occurs for each employment. This means that employments that came off the scheme and were later re-furloughed are only included in this analysis once, with the most recent furlough end date determining the last reported furlough month.

We use the dataset produced for the Earnings and Employment statistics to determine whether an employment is payrolled in any given month; this data includes probabilistic imputed values to improve the data quality, especially for more recent months. Employers do not always submit PAYE returns on time and do not always provide leaving dates for ended employments, so the probabilistic imputed values indicate the probability that an employment is still payrolled where there is no PAYE return nor a leaving date. For detailed information, see the methodology document for the Earnings and Employment statistics.

2. Limitations

• Employers had until 31 July 2020 to submit claims for employments furloughed to 30th June 2020, meaning only claims data for employments furloughed on or before that date are complete. Employments last reported furloughed in June might be revised downwards in future releases, as employers can still submit July furlough claims. Trends in past submissions suggest that over 90% of July furlough claims were submitted by the date of this data extract (31 August 2020). Therefore, we expect the numbers for furloughs last reported in June to be revised downward by ~10% when we receive all the July furlough claims (deadline November 2020). This will affect the numbers in both Tables 1 and 2, but we expect the proportions in Table 3 to be relatively unaffected (as both numerator and denominator will drop).
• Payrolled redundancy payments are paid via PAYE, so redundancy pay will appear as an employment. This means we will see a larger number of employments still on payroll than actual continued employments; we might expect a single redundancy payment to be paid via PAYE the month after the last furlough period, but for no further payments to be made.
• Quarterly paid employments are mostly captured in this release, as payments for the first quarter (Apr-June) have been received in RTI, resulting in a probabilistic imputed value of 1. Annually paid employments will mostly have a probabilistic imputed value of less than 1, resulting in the data showing fewer currently employed than we expect.
• The RTI probabilistic imputation methodology extrapolates the likelihood of an employment still being on payroll from all past employers’ submission behaviour. When using this imputation methodology, we assume that employers who have furloughed staff (whether the staff are currently employed or not) do not differ behaviourally in submitting PAYE claims from the entire population of employers.