Selected Subjects: Fertility of Marriage

Next Selection Previous Selection

4.11 Fertility of Marriage

4.11.1 1911 - 21

From 1861 onwards it was usual to find, in those parts of the Reports which commented on analyses derived from returns made on civil condition, reference to an average annual fertility rate based on a comparison of the number of enumerated wives who had not completed their forty-fifth year with the average annual number of legitimate births registered in specified years. The General Report (p xvi) on the census of England and Wales in 1871 contains an interesting comparison with comparative fertility rates in France at that time.

At the census of 1911, however, a direct enquiry into the fertility of marriage was made for the first time. To obtain the necessary particulars four additional columns (6-9) were added to the schedule as follows:


Write "Single", "Married", "Widower" or "Widow" opposite the names of all persons aged 15 years and upwards State for each Married Woman entered on this Schedule the Number of:
Completed years the present Marriage has lasted. If less than one year, write "under one" Children born alive to present Marriage (If no children born alive write "None" in Column 7)
Total Children born alive Children still living Children who have died
5 6 7 8 9

In Scotland the final column was omitted. It should be noted that widows were excluded from the scope of the enquiry which was restricted to the fertility of existing marriages.

There was little evidence to cause doubts about the reliability of the returns as to the numbers of children since they appeared to be reasonably distributed. In the statement of marriage duration, however, there was a strong tendency to concentration on round numbers - 10, 20, 30, 40, as met with in the statement of ages. Attention was drawn to the significance of this in the report Fertility of Marriage. The age at marriage was derived by deduction of the duration of marriage from the age at the census. Thus the deduction might well have been based on a faulty statement as to duration. It also included in many instances persons nearly a year younger than the stated figure. For example: a statement on the schedule that a woman is 31 years old and married 6 years gives a resulting age at marriage of 25 years. At the census she may have been 31 years exactly or 31.9 years; similarly, she may have been married 6 years exactly or 6.9 years; if she was 31 years exactly and married 6.9 years, she was actually 24.1 years old at marriage; while if she was 31.9 years and married 6 years exactly she was actually 25.9 years old at marriage. Accordingly, age at marriage 25 could include any age over 24 and under 26.

In the case of England and Wales the analyses were limited to the fertility of 6,014,319 out of the 6,630,284 married women enumerated at the census, or 90.7 per cent of the whole. The balance consisted of 493,679 wives whose husbands were not enumerated on the same schedule and of 122,286 rejected because of obvious defective information, eg in the case of a woman of 21 married for one year and returned as the mother of seven children, who were presumably born to the husband by a previous wife. It was felt that a slight reduction in the size of the sample of population was preferable to a more complete sample which included defective information. For similar reasons the fertility study in Scotland was restricted to an analysis of the returns made in respect of 680,684 wives out of a total of 762,835 enumerated, ie nearly 90 per cent.


The tabulations involved were very lengthy. The tables for England and Wales were made available in two parts of Volume XIII in the series of reports. Part I of Fertility of Marriage was published in 1917. Owing to the war Part II , which contained a report as well as tables, was not published until 1923. In Scotland the detailed tables were not published but it was stated in Volume III of the Report on the 1911 Census of Scotland (p xxv), in which only abstracts were published, that 'The original tables, though unpublished will be preserved in the Registrar General's Department, and will be available for purposes of statistical study to any interested in them'.

In view of the considerable material derived from this enquiry in 1911, it was not deemed necessary to repeat it in the census of 1921, and that of 1931 was restricted in scope in any case, partly by reason of the need for economy and partly because the expectation at the time was that the next census would be in 1936. Certain fertility statistics, derived from the analysis of data on dependency, were published in the Reports on the census of 1921.

4.11.2 1951 - 66


For the 1951 Census this topic was reinstated after an interval of forty years apart from some fertility information collected in 1921 which was incidental to the preparation of the tables on dependency and orphanhood, otherwise very little current information on fertility was collected until the Population (Statistics) Act, 1938 greatly extended the data obtained at birth registration. When the Royal Commission on Population was appointed in 1944 they found it necessary, in order to discharge their task, to take a special Family Census. Their enquiry, in February 1946, covered a ten per cent sample of women in Great Britain who were or had been married and the preliminary results greatly contributed to the Commission's conclusions; the final report was published in 1954.[i] Because the enquiry was confined to fertility questions, it was possible to ask for more details than in a general population census, and in particular for the date of birth of each child. On the other hand the Family Census unavoidably had to be conducted in such a way that cross-tabulation of fertility characteristics with others such as occupation could not be made in any detail, so that differential fertility analysis was severely restricted and had to be made in units not strictly comparable with those identified in the general census of population.

The enquiry took place before many of the births postponed during the war had been 'made up' but by 1951 this was for all practical purposes complete, so that the fertility recorded up to census date would be expected to present a more normal picture than it had done in 1946. New census infor¿mation was also needed to provide a firm base, after the disturbed war and post-war years, for the annual estimates used to calculate fertility rates and mean family sizes from birth registration statistics.

The questions

The questions in the private household schedule bearing most directly on fertility, in addition to those on sex, age at census, and whether single, married, widowed or divorced, were as follows:

State for each married woman
under age 50 included in this
born in
The month and year of her present marriage at (a).

If married more than once, state also the month and year of her first marriage at (b).
The total number of children born alive to her (including any of a previous marriage and any that have died).

If none, write "none".
Whether she has given birth to a live-born child during the last twelve months, ie on or after 9th April, 1950 ("Yes" or "No").


They applied only to women, as is usual in enquiries of this sort, because biologically fertility is more closely dependent on such characteristics as the age of the mother than of the father; but where the husband was enumerated on the same schedule, and the woman had only been married once, his characteristics, such as age and social class, could be combined with hers in the tabulations where needed.


The questions were also confined to married women who were under the age of 50 at the time of the census and to reduce the work caused by the extensive tabulation programme the analysis was carried out on two samples. A one-fifth sample of married women under 50 was chosen for the general analysis and a four-fifth sample of those aged between 45 and 50 for the analysis of completed fertility. An exception was made in the case of women married more than once and they were tabulated in full. The resulting data were big enough but there were bound to be cells with small frequencies especially at the edges of the distributions and in highly cross-classified tables.

Standardised Ratios

In the first three sections of the main tables, standardised ratios have been shown. These are the ratios of the overall rates (mean family size, etc) which would have occurred in the basic population with which comparison is made if it had been subject to the detailed rates of the sectional experience, to the overall rates which actually occurred.

Rates and ratios based on small frequencies have been printed with signifi¿cance indications. For details see Ferti1ity Report, page x.

Differences between 1911 and 1951

The 1911 Census confined its question to marriages existing at census date and to children of those marriages. It did not take into account, therefore, the children and the duration of any previous marriage. Consequently in 1911 the total number of children born to the total number of married women would be understated by the number of children born of previous marriages. So also would the total marriage duration since the periods of all previous marriages were left out of account. The 1951 Census, on the other hand, asked the date of the first as well as that of the existing marriage and for particulars of children of all marriages. What it did not do was to establish the duration of each marriage where the women had been married more than once. Consequently, the total number of children cannot be accurately related to the total duration of marriage of women in this category as the interval when they were widowed or divorced is not known.

Comparison with 1911

It is difficult to make any direct comparison between the 1911 and 1951 statistics. Reference has already been made to the differences in the questions put at the two censuses; also, the data produced for 1911 while not limited to women under 50 at the time of the census excluded women over 45 at the time of marriage and women not enumerated in the same household as the husband. Women aged 45 and over at the time of marriage were excluded on the assumption that such marriages were likely to be infertile and women not enumerated with husbands were excluded as the analysis involved the age at marriage of both husband and wife, and this could only be obtained where they were living in the same household. The 1951 data on the other hand, while relating only to married women under 50 include, unless otherwise, indicated, all married women in the age group with no exceptions.

Quality of response

There is evidence to show that premarital births were included in the answers to total children born alive to married women, and steps were taken to exclude apparent inconsistencies; also there were mis-statements associated with premarital conception although they appear less important and to vary less consistently with social class than in 1911.

The most serious deficiency was a failure to state the date of first marriage for women who had been married more than once. As a result, women for whom the information was not given appeared to have been married once only which seriously distorted many of the census figures.

The existence of this problem was brought to light by the analysis of the fertility data from the One per cent Sample. It was clear that, in spite of the tiny fraction of the total number of married women affected, the bias imparted to the data was too great to be tolerated. It was therefore decided to adjust all the raw data on the following assumptions:

  1. that the proportion remarried among all women enumerated who entered their current marriage at a given age and in a given calendar period, was the same as that shown by the original marriage registration statistics;
  2. that the family size distribution and current fertility of women whose remarriage had not been stated was the same as that of women known to be remarried who had entered their current marriage at the same age and in the same calendar period;
  3. that the distribution of the 'not stated' cases by duration of (and hence also age at) first marriage was the same as that of the stated cases for any given census age and family size, irrespective of date of current marriage.

One Per Cent Tables

The One Per Cent Sample Tables are based on a much smaller sample and are subject to their own sampling variations but the chief difference between them and the main tables is in the later adjustment detailed above. It means that at marriage ages over 30 (or, at longer durations, 35) the data are seriously biased and must be used with great caution, if at all.

Differences in Scotland

In Scotland no adjustment was made to the figures of remarried women. It was considered that the number of such cases, if known, would not be likely to have any material affect on the accuracy of the Scottish data.

Publications - England and Wales

The main Fertility Report contains a comprehensive commentary and tables. The tables are divided into five sections A to E, of which D and E use, in addition to census data, material from the Family Census of 1946 and from the annual vital registration statistics. The only other references are in the One Per Cent Sample Tables and the General Report .

Publication - Scotland

The subject of fertility is contained in Volume V and the tables are similar to those for England and Wales.


The 1961 Census produced two fairly basic changes in the questions as set down on the schedules. Firstly the scope of the questions was widened to include all women of any age who were or had been married. The other change was to ask all widows and divorced women, or women married more than once to give the date when the first or only marriage ended. The remaining questions were unaltered but the order was changed. As the question covered all ever-married women the inclusion of the women aged 50 and over at census date permitted the analysis to be extended to a wider group, that is women with uninterrupted first marriage. These were women whose age at their only marriage was less than 45 and whose marriage was still in existence at census date, plus any other women whose first or only marriage took place before they were 45 and lasted until they were 45, irrespective of their marital condition when the census was taken. Certain tables in the 1961 Fertility Tables still related to women married once only, but these women would be of any age at census though they must have been under 45 when they married.


Standardised ratios were again produced as in 1951 and full details are given in the Fertility Tables .


The procedure on sampling was dictated by the introduction of the ten per cent element in collecting information on a number of subjects which rather cut across some of the data used in fertility tabulations. Consequently some tabulations are based on the full count and others are derived from information given in the ten per cent sample.

Sampling Error

In other subjects sampling error has only affected figures derived from the ten per cent sample. For fertility however indications of significant sampling errors have been attached to mean family sizes, fertility rates and proportions infertile derived from both the full count and from the ten per cent sample. This indication appears where such a statistic or a ratio of such statistics is based on a relatively small number of women. This shows that although the statistic concerned may be based on a full count of the number of women included in a sub-group of the population (eg the women with a particular number of children), it will not provide a reliable estimate of the true fertility of this particular group. Suppose, for example, that the underlying pattern of fertility for a given group of women is for their mean family size to rise with increasing marriage duration, then for those durations where the census counted large numbers of women, the underlying pattern should be clearly reflected by the census count. If the census only counted one or two women at certain durations then the census count could well depart from the underlying pattern because of the characteristics of the one or two women at those durations. The variance or relvariance (ie the variance of a rate, etc divided by the square of that rate, etc) provides a measure of this unreliability.

Two sorts of quantity, totals on the one hand and statistics such as mean family sizes on the other appear in the tables. They can both be regarded as subject to chance variation in the way described above. This element of variation is to be distinguished from that due to bias. Where the derived statistics were based on the ten per cent sample the computed relvariances were multiplied by the factor of 0.9.

Bias in the sample

When the ten per cent sample was checked for bias it was found that there was considerable under-representation of one person households and of large households. For any stated size of household there is a clear gradient from too few households occupying few rooms towards too many households with large numbers of rooms.

The effect of this bias on those fertility rates, etc based on the ten per cent sample cannot be derived directly from a comparison of any of the published tables, but it seems likely that any affect of bias has not been enough to disturb the fertility pattern.

Other evidence of bias

There is however some evidence of bias in the number of women excluded from the ten per cent sample and full count fertility tables because of the failure to answer the relevant questions. Comparison has shown that the ten per cent sample was biased in that it included 20 to 30 per cent more married women with fertility data not stated than it should have done. However, as women with not stated fertility data represented only three or four per cent of all married women, the bias in the number of women with stated fertility data was only about one per cent. The excess in the sample of widowed and divorced women with fertility data not stated is much smaller.

Quality of response

As well as comparison of the full count and ten per cent sample the post-enumeration survey was used as a measure of quality, and there is no evidence of significant mis-statements other than in the number of live-born children.

In processing the census results it was found that there were a large number of women who failed to answer the questions or who put a dash or a stroke in the answers relating to number of children. The women failing to answer were excluded from the tabulation but from a provisional comparison with the post-enumeration survey it was decided to convert dashes or strokes to 'None1 or 'No' in answer to number of children and to whether the women had had a child in the last year. The fact that so many women failed to answer is probably a reflection on the wording of the question which failed to give specific instruction to childless women. The corresponding question in 1951 had included an instruction that childless women should answer 'None'. The subsequent final analysis of the post-enumeration survey indicated that only four out of every five should have been so treated and it follows that the number of childless women has been slightly overstated in the tabulation and for women with uninterrupted first marriage the over-statement can be estimated at 6.5 per cent and such over-statement, though not large, is certainly not negligible.

Publication - England and Wales

In addition to the main volume, figures appear in the Great Britain Summary Tables and in the General Report .

Publication - Scotland

The main tables appear in the Fertility Tables and are much more comprehen¿sive than those published in 1951.


There were no questions on fertility in the sample census of 1966 but tabulations showing numbers of children in families appear in the Household Composition Tables and tabulation by type of marriage and numbers of children in the Commonwealth Immigrant Tables .

[i] The Trend and Pattern of Fertility in Great Britain. A report on the Family Census of 1946. By D V Glass and E Grebenik. Papers of the Royal Commission on Population, Volume VI, London, Her Majesty's Stationery Office, 1954.

Office of Population Censuses and Surveys/General Register Office, Guide to Census Reports: Great Britain 1801-1966 (London: HMSO, 1977) Crown Copyright. The Office of National Statistics has granted the Great Britain Historical GIS Project permission to computerise this publication and include it in this web site. All other rights reserved.

Next Selection Previous Selection