CAIN Web Service
ESRC Data Archive Bulletin:
Family Expenditure Survey
Text: Pat McGregor and Patricia McKee ... Page Compiled: Fionnuala McKenna
Material is added to this site on a regular basis - information on this page may change
Family Expenditure Survey
Pat McGregor and Patricia McKee
University of Ulster
The Family Expenditure Survey (FES) in Northern
Ireland (Archive study number 33240) is the responsibility
of Policy Planning and Research Unit (See Jardine).
In 1967 the Northern Ireland sample was collected separately from
the UK for the first time. The sample was increased to 912 from
about 130 to facilitate separate analysis and the exercise was
supervised by the then Ministry of Finance. Thereafter a random
sample of the Northern Ireland sample was forwarded to the UK
The 1967 NI sample design was a two stage
one with stratification by area and population density. Three
broad geographic divisions were employed; Belfast,
the western counties and the remainder. Outside Belfast urban
and rural areas were identified by the local authority classifications
and the required number were selected randomly with a probability
proportional to their size. From each area selected in the first
stage, together with the Belfast County Borough.
addresses were selected from the list in the Valuation Office
using a random start and constant interval for each area,
with the number of addresses varying between strata - the 1967
Report gives further details.
Although local government was reorganised
in 1973, the above sample design continued to operate until 1975.
Thereafter the electoral wards of the 26 District Councils became
the primary sampling units (psu) of the new design, which again
was two stage and stratified by population density and area (the
geographic stratification was similar to the earlier design).
Wards within each of the three geographic
strata were subdivided into three equal groups based upon a population
density criterion, giving nine strata in all. The first stage
of the sampling consisted of the selection of wards with a probability
proportional to size. From each of the Belfast strata, two wards
were chosen per quarter, against six for the remaining strata.
The second stage of sampling again used the Valuation List, with
the number of addresses varying between strata - details are given
in the 1976 Report. The current NI design, adopted in the mid
1980's, maintains the geographic stratification but now consists
of a simple random sample within each strata with 20% of addresses
drawn from the Belfast stratum, 45% from the East with the remainder
from the West.
In Great Britain, the FES (Archive study number
33057) sample design was and is significantly different, reflecting
the considerably greater population. Up to 1985 a four stage,
stratified rotating design with a uniform overall sampling fraction
was employed. The local authority areas of GB were stratified
into (i) 16 standard regions, (ii) type, based on population density
and (iii) economic character, based upon rateable value.
Each PSU selected was used for 4 quarters
with one quarter of the total PSU's being reselected each quarter.
(More details are given in Kemsley et al. 1980).
Along with other household surveys conducted
by the Office of Population Censuses and Surveys. the FES in GB
changed its sample design in 1986 and began using the small user
postal address file with the postcode postal sector as its psu
instead of the electoral ward (see Wilson and Elliot, 1987). Some
672 postal sectors were randomly selected during the year after
being arranged in strata comprising standard region, area type
and the proportion of owner-occupiers and proportion renters according
to the 1981 Census. The result was a less clustered sample design
which it was hoped would increase the precision of estimates.
Clearly the contemporary GB and NI FES employ
a radically different sample design. The extent to which this
affects the relative accuracy of the two surveys would be a research
project in itself. However, a more pressing issue is provided
by the response rates to the two surveys, which are displayed
in Fig 1. The UK response rate has fluctuated about 69% whereas
in N. Ireland the response rate steadily declined from 80% in
1967 to 53% in 1989. This could be related to the political situation
- indeed in the years 1970-74 a total of 183 addresses
were abandoned due to civil unrest.
A variable response rate naturally exacerbates
the problems associated with the differential response rates by
social and economic category. UK response rate rose from 67% in
1980 to 72% in 1981. As a result the proportion of households
headed by a person of 30 years and under 50 rose by 1 percentage
point (Employment Gazette, Sept, 1982). If the 1980 Survey were
reweighted by the 1981 figures the result would be a 0.8% increase
in the total expenditure, though only a 0.1% increase in total
expenditure per person. (Employment Gazette, Dec, 1982).
With the smaller sample size in N. Ireland,
effects can be more dramatic. In 1980 (as in 1978) there were
about 1.5% more pensioners than might be expected, thus depressing
the estimates of average household size, income and expenditure
- the latter two were possibly 4% higher than the 1980 FES had
suggested (NI FES Report for 1980).
Turning to the questionnaire itself, the NI
FES uses the same forms as GB except, from April 1989, for those
parts covering rates and the community charge. A significant development
in NI was the addition from the 1988 survey onwards, of a voluntary
question concerning the religious denomination of the respondent.
In Table I the residual responses include other religious denominations
as well as those who refused to answer the question,
so it would appear that whatever the reticence to cooperate with
the FES overall, this question elicits a strong response.
Religious Denomination of Heads
|Catholic ||218 (36.0%)
The FES data for Northern Ireland are deposited
in the Archive. There are special access conditions which require
that details of the proposal research be submitted. On receipt
of approval the data is forwarded to the researcher in either
SIR format or, if required, in SPSS format. The FES data are structured
with households at the top of the hierarchy, individuals at the
base and either tax units or benefit units in between.
CSO Annual Abstract of Statistics,
CSO, Family Expenditure Survey, Annual Reports.
Dept. of Employment, "Pattern of household
spending in 1981", Employment Gazette, Vol.90,
Dept. of Employment, "Household spending in 1981",
Employment Gazette, Vol.90, Sept.1982,
Kemsley, W.F.F., Redpath, R.U., and Holmes, M. (1980), Family
Expenditure Survey Handbook, London; HMSO, 1980.
NI Dept of Finance and Personnel, NI Annual Abstract of Statistics,
NI Department of Finance and Personnel, FES reports: 1967-81
annual, since then PPRU Monitor No 3/88; No 2/89; No 1/91.
Wilson, P.R. and Elliot, D.J.
(1987) "An evaluation of the Postcode Address File as a Sampling
Frame and its Use within OPCS", Journal of the Royal
Statistical Society, Series A, Vol.150, Part
Return to Contents