(i) Sampling Method

The sample of 12,000 households set as a target was stratified by civil defence region and size of town on the basis of the Registrar General’s population figures for March 1945. For example, 16% of the population lived in the North-Western Region, and 16% of the sample of 12,000 was allocated to that region. Within the North-Western region that proportion of the sample was then allocated according to the percentage of persons living in towns of four different sizes - up to 40,000 inhabitants; over 40,000 to 80,000 inhabitants; over 80,000 to 200,000 inhabitants, and over 200,000 inhabitants. One or more towns were chosen in each group of towns of the given size. The towns were chosen on the basis of geographical position within the region. A list of the town visited is given in Appendix C.

It will be noticed that although it was proposed to sample households the interviews were allocated on the basis of population. No other method was possible at the time, since the number of households in each region was then unknown. The average family size does not vary greatly from one region to another, however, and a comparison of the proportions of the sample allocated to each region with the proportions which would have been allocated if the household distribution had then been known, shows very slight differences between the two. The North-West, for instance, where 16% of the sample was allocated should have had 15%.

The final selection of addresses to be visited was made from the Rating Lists of the Local Authority. Addresses were drawn at regular intervals from these lists, and it was laid down that in the first place the housewife at this address was to be interviewed and the details about members of her household obtained from her. In cases of losses due to empty houses, the housewife not being at home, or for other reasons, a list of substitute addresses was drawn in the same way.

A complication of this method of drawing addresses is the possible presence of two or more households at each address. In certain instances it is possible, therefore, that areas with a high proportion of multiple tenanted dwellings will be over represented in the sample. Regionally, this complication is not of great importance outside London, since an analysis of the figures on multiple tenanted dwellings derived from this sample shows in 10 regions an average of 95% of dwellings with one household, with a variance of 1%.

Sampling losses have introduced greater variations between regions than would have been caused by the presence of multiple dwellings, as the note on Regional analyses points out later.

(ii) Sample Analyses and Comparisons

Because of a limit on the length of time in which the work could be completed the sample finally obtained was reduced to 11,276 households. Included in these households were 40,504 persons, of whom 35,941 lived at home at the time of the inquiry, the majority of the remainder being in the Armed Forces. This sample was six per cent smaller than the number of addresses taken, and, allowing for multiple dwellings at these addresses, 13% less than might have been expected. Analyses of the sample and comparisons of the data with material derived from other sources suggest, however, that the affect of this loss of interviews upon the validity of the data is not great. A comparison is made below of the age distribution of the sample of persons at home at the time of the inquiry with the age distribution of the civilian population as calculated by the Registrar-General for June 1945.

The differences between the percentages derived from the Survey and the Registrar-General’s returns are in all cases small. It must be remembered also, that the Registrar-General’s figures include persons in institutions. The sample figures do not.

Further analyses of the sample, where possible with comparisons from other sources are given below:-

8

(iii) Sub-Sample: Persons over 60 in employment; Mid 1945

The names and addresses of all old persons in employment were recorded and they were revisited in August 1945. Altogether, 1,571 old persons were recorded as being in employment, and of these 961 were interviewed in August. The loss arose principally because of the low rate of interviews per day, and a limit on the time in which the interviews could be completed, an unexpectedly high refusal rate, (possibly due to re-interviewing and certainly partly due to suspicion on the part of old persons about inquiries into their work - a reflection perhaps of a sense of insecurity) and partly to removals to unknown addresses.

Fortunately it is possible to compare the sub-sample obtained with the original sample. Comparative age distributions follow:-

The principal differences are a slight loss of persons aged 60-64 years and a slight gain of persons 65-69 years.

The proportion of men and women are the same in both samples.

An occupational distribution of the original sample and the sample obtained shows slight differences.

In the sub-sample obtained there are somewhat higher proportions of labourers end operatives than in the original sub-sample.

The proportion of part-timers and full-timers are almost identical.

An analysis by economic group reflects to some extent the occupation differences between the two samples.

The sub-sample obtained has, therefore, a slightly higher proportion of older persons, labourers and operatives, and persons in the lowest economic group, than the original sub-sample.