To ask the Chancellor of the Exchequer what analysis he has made of whether the assessment of earnings levels for (a) Worcestershire, (b) Dudley, (c) Walsall and (d) Sandwell in the New Earnings Survey are reliable and stable. 
I have been asked to reply.The criteria used by the Office for National Statistics to establish whether estimates of gross weekly pay from the New Earnings Survey are sufficiently reliable to publish are that the sample response should be at least 30 and the relative standard error of the mean should be less than 5 per cent. The published estimates for Worcestershire, Dudley, Walsall and Sandwell fully meet these criteria, as indicated in the following table.
|Sample sizes and relative standard errors for New Earnings Survey estimates of mean gross weekly pay1 in April 2002|
|Sample response||Relative standard error of the mean (Percentage)|
|1 For full-time adults whose pay was unaffected by absence during the pay period.|
New Earnings Survey, ONS
In terms of the stability of New Earnings Survey estimates over time, the survey is based on a 1 per cent. sample of employees comprising all those whose National Insurance numbers end with a specified pair of numbers (the same pair of numbers has been used since 1975). Those individuals for whom completed questionnaires are received in successive surveys constitute about 80 per cent. of the total. As such, the estimate of the change in average weekly pay between successive years is based largely on the changes in pay for individuals who were in the sample in both years.
However, for small areas, the stability of estimates of mean pay over time can be more volatile, reflecting in part the underlying volatility of earnings, but also the variability associated with sampling.