New Earnings Survey

One of the short comings of the 1991 Census is the lack of any information on income, the situation for 1991 can be partially remedied by using mean earnings information from the New Earnings Survey (NES) to attach to individual in the SARs

Mean hourly earnings (excluding overtime) have been derived from the New Earnings Survey for 1991 at the level of SOC minor (NESSCORE) , the percentage standard error of the mean (NESSTERR) and the number of cases on which the information is based (NESNUM). Earnings are broken down by sex, full-time or part-time work, age group and living in South East or Rest of Great Britain.

The NES is a large survey (sample size about 160,000) conducted annually by the Office for National Statistics and collects information from employers about the earnings of their employees. The data are collected under statute and only aggregate (tabular) data are released.

We are grateful to the Department of Employment for allowing the use of these data and in particular the staff at Runcorn for extracting the relevant tables.

In order to attach aggregate earnings information to individuals in the SARs, hourly earnings (excluding overtime) in 1991 were obtained in aggregate form broken down by the following

characteristics:

Geographical groupings:

London and the south-east; rest of GB

Sex:

male; female

Employment status:

Full-time; part-time

Age:

16-19; 20-29; 30-49; 50+

Occupation:

level of minor SOC- 73 categories (The categories of minor SOC are shown as a derived variable in the SARs codebook.)

We have used hourly earnings excluding overtime because we take the value to represent the earnings potential of someone in the relevant occupation and in the particular age, sex, employment status and regional group. There are many other factors, including amount of overtime worked, that affect actual individual earnings.

Where there are fewer than 3 values in an NES category there are no data shown. In an attempt to minimise missing values, a coarser breakdown of sub-major SOC groups have been used for men working part-time. NES values do not apply to the self-employed and scores have not been added to those on Government schemes, nor for the armed forces (parts of Minor SOC 15 and Minor SOC 60).

There are four variables concerned with the New Earnings Survey:

NESSCORE: NES Occupational Score

This is assigned to persons with econprim=1 or 2.

NESNUM: Number in sample

NESSTERR: NES Standard error

NESNOTWK NES not in work Score

Assigned to person who have Econprim > 5 who have an occupation stated and who were not self employed.

For part-time employees it is important to note that the New Earnings Survey is likely to over-estimate earnings as the sampling frame for the survey is based upon National Insurance number and those earnings below the Lower Earnings Limit (LEL) - #52 in 1991 - do not pay NI and will generally not be included in the sample (some employees who do not pay NI will be included in 'centralised returns' from large organisations). However, this bias is reduced by using hourly earnings rather than weekly earnings.

The NES information has also been added to people not currently in work, using information on occupation and employment status related to their last occupation, (employee/self-employed status from the SEG of the last occupation) but age and geographical located to the 1991 situation. To avoid confusion is has been set up as a separate variable - NESNOTWK. Note, however, that the 'number in sample' and standard error will be the same as for NESSCORE.

Generally, individuals currently in employment and who report an occupation have been successfully allocated NES values. Only for men working part-time are more than one per cent of individuals without a valid NES score.

Those men without a valid NES score on the 2% SAR (674 full-time and 585 part-time) are fairly evenly distributed throughout the occupational spectrum. The women without a valid NES score are disproportionately found in Minor SOC 33 or 39 (combined in the 2% SAR).

For those not in employment (ECONPRIM 6-10), and aged 16-60, the percentages without a NES score are much higher and reflect the fact that people in this group are much less likely to report a previous occupation. For men, 54.6% have a NES score whilst for women this falls to 46.6%.

Use of mean income scores

Work conducted in the mid 1980's by Stewart (1983), Nickell (1982), and Metcalf and Nickell (1982) showed the utility of mean earnings values attached to individual level data as an indicator of occupational attainment. There is also substantial literature in the US which uses aggregate income values in this way (eg Roos and Hennessy, 1987; Jasso and Rosenzweiget, 1990) and recent work using NES data with the SARs (Ladipo, 1995).

How do NES scores relate to actual earnings?

It is obviously important to remember the NES scores do not reflect the amount earned by the individual in question. Earnings are influenced by a range of variables including size of firm, number of years experience, qualifications, differing grades within an occupation, and the local labour market. There is also likely to be variation by ethnic group. None of this variation is captured by using mean values which reflect only differences by broad occupational grouping, age-group, sex, full or part-time working and whether living in the south-east or not. Recent work using longitudinal data from the National Child Development Study (Dale and Egerton, 1995) demonstrates the importance of a much larger number of variables in explaining actual hourly earnings.

In order to establish in more detail the way in which NES scores relate to actual earnings and also to other measures of occupational attainment we have added the NES scores to individuals in the 1991 General Household Survey, using the same matching criteria as for the SARs. This then allows a direct comparison on aggregate income scores and 'real' earnings.