Office for National Statistics Low Pay Methodology
1 Prior to 2004, due to the limitations of the surveys available, the Office for National Statistics (ONS) used a central estimate methodology to produce its low pay statistics. Under this methodology the findings of the New Earnings Survey (NES) and the Labour Force Survey (LFS) were averaged to produce the low pay estimates (the central estimate). The NES, although providing reliable information from employer records, was likely to under-sample those on low pay, especially if they earned less than the Pay As You Earn (PAYE) threshold. The LFS, on the other hand, would cover such workers, but it was self-reported and contained many proxy responses - both factors making for an element of uncertainty.
2 The NES was replaced by a new annual earnings survey, the Annual Survey of Hours and Earnings (ASHE) in April 2004. ASHE, like NES, is based on a one per cent sample of employees. Information on earnings and hours is obtained in confidence from employers. The survey does not cover the self-employed. In 2004 information collected on earnings and hours related to the pay period which included 21 April 2004. The earnings information relates to gross pay before tax, national insurance or other deductions, and excludes payments in kind. It is restricted to earnings relating to the survey period and so excludes payments of arrears from another period made during the survey period. Any payments due as a result of a pay settlement but not yet paid at the time of the survey will also be excluded.
3 There are four main differences between ASHE and NES.
- First, ASHE results are weighted to the number of jobs given by the LFS, which itself is weighted by the Census and mid-year population estimates. This weight differs slightly from that previously used for the low pay statistics. NES, being a random one per cent sample, was unweighted for the main earnings statistics.
- Second, unlike NES, ASHE imputes for item non-response.
- Third, the median replaces the mean as the headline statistic. The median is the value below which 50 per cent of employees fall. It is preferred over the mean for earnings data as it is less influenced by extreme values and because of the skewed distribution of earnings data.
- Last, and perhaps most importantly for the low pay statistics, the coverage of employees for ASHE extends that of the NES.
4 The sample has been boosted in order to improve the coverage of those at the lower end of the earnings distribution. A supplementary survey now collects information on those businesses which are registered for Value Added Tax (VAT) but not registered for PAYE. In addition, follow-up surveys have been introduced for those changing jobs, or new entrants starting jobs, between ONS conducting the initial PAYE sample in January and the survey reference period in April.
5 People working for small firms and those who move jobs frequently are more likely to be low paid than others and these additional samples therefore improve coverage of the low end of the pay distribution. The improved quality and coverage of the ASHE data has enabled ONS, from April 2004, to replace the former central estimate methodology for its low pay statistics with data drawn from ASHE alone.
6 The NES data has been revised from 1998 to 2003 to take account of the imputation and weighting used in ASHE. Unfortunately, the extended coverage is not available for NES. The revised imputed and weighted NES is now known as ASHE without supplementary information. The ONS has therefore revised its low pay estimates for 1998-2003 using a central estimate of LFS and ASHE without supplementary information. ASHE with the extended coverage (as well as the imputation and weighting) available from 2004 is known as ASHE with supplementary information.
7 Overall, ONS concludes that there is little difference between the old central estimates calculated using the old methodology and the results gained using the new ASHE methodology. This may be true overall, but we have found that the new ASHE survey brings significant improvements to calculations relating to the low paid.
8 The results of the new methodology are shown in Table A7.1. This shows the number of jobs that ONS now estimates to have been paid below the minimum wage between 1998 and 2003.
Table A7.1 ONS Revised Estimates of Employee Jobs Paid Below the National Minimum Wage for Those Aged 18 and Over, 1998-2003
Source: ONS estimates based on a central estimate of the LFS and ASHE without supplementary information, Spring 1998-2003.
Note: Figures for Spring 1998, before the National Minimum Wage was introduced, are for the number of jobs paid at less than £3.00 per hour (aged 18-21) or £3.60 per hour (aged 22 and over).
9 We can see how these new estimates differ from the old methodology in Table A7.2. This shows the latest estimates of those paid below the minimum wage using the LFS, NES, the central estimate and ASHE. The Table demonstrates that the estimates have become much closer over time. The ASHE estimates in 1998 are much lower than the LFS, NES and central estimates.
Table A7.2 Comparison of Old and New ONS Methodology for Estimates of Employee Jobs Paid Below the National Minimum Wage for Those Aged 18 and Over, 1998-2003
Source: ONS estimates based on LFS, NES, central estimate and ASHE, Spring 1998-2003.
1. LFS using revised weights consistent with the population estimates published in Spring 2003.
2. NES using annual revisions and Spring 2003 population weights.
3. Central estimate is the average of LFS and NES.
4. Using the new ASHE methodology with weights for low pay data.
10 Further details of the new low pay estimates for 1998-2004 can be found in Labour Market Trends (Bird (2004), Daffin (2004), Milton (2004)) and on the National Statistics website at:
For comparison, estimates are also given for the old low pay methodology using a central estimate of the NES and LFS at:
The new low pay methodology is explained in more detail at:
and the new ASHE methodology at: