Full text of DOI:10.1111/1468-0289.00185

Economic History Review, LIV, 1 (2001), pp. 87–107
Women’s work in census and
survey, 1911–19311
By TIMOTHY J. HATTON and ROY E. BAILEY
t has long been believed that women’s work has been undercounted
in the censuses of the late nineteenth and early twentieth centuries. If
I
so, then the ‘revolution in women’s work’ as reflected in the rise in the
labour force participation ratio from the middle of the twentieth century
might be less dramatic than census statistics suggest. A number of
observers have argued that some important classes of women’s market
work were underreported by the respondents and either misclassified or
ignored by the census takers. However, these evaluations are often based
either on assumptions about under-enumeration or on very fragmentary
data. One alternative source is the relatively comprehensive surveys of
working-class living standards, a number of which were undertaken from
the turn of the century onwards. This article compares census partici-
pation rates for certain towns with statistics for women’s work derived
from social surveys. The comparisons suggest that, if anything, labour
force participation rates derived from the census overstate the ‘true’
participation rates. Since the under-enumeration of women’s work in the
early twentieth century is largely a myth, the subsequent rise in partici-
pation rates is not overstated; in fact it may be slightly understated.
I
The course of women’s labour force participation rates in the twentieth
century from the decennial census of England and Wales is plotted in
figure 1. These graphs show the well-known pattern of participation rates
beginning a steep rise somewhere in the middle of the century. The
graphs for women aged 15–59 and for married women show a sharp
break from an essentially flat trend up to 1931 followed thereafter by a
steeply rising trend. The levels and trends for the nineteenth century are
more difficult to discern because of changes in classification and in
methods of enumeration, but a number of writers have argued that
participation rates declined gradually to reach a low point somewhere in
the late nineteenth century. In the period 1891–1931, recently dubbed
1 The research for this article is supported under the ESRC ‘Future of work’ Programme grant
no. L212252004. The computerization of data used here from the New survey of London life and
labour was supported by ESRC grant nos. R000235697 and R000221981. For the computerization
and coding of the data we acknowledge the collaboration of the other members of the project team,
Dudley Baines, Paul Johnson, and Anna Leith. We have benefited from valuable and constructive
comments on an earlier draft from Edward Higgs, Sara Horrell, and Matthew Woollard. We are
also grateful for useful comments from anonymous referees.
Economic History Society 2001. Published by Blackwell Publishers, 108 Cowley Road, Oxford OX4 1JF, UK and 350 Main Street, Malden,
MA 02148, USA.

88
t. j. hatton and r. e. bailey
80
70
60
50
40
30
20
10
0
Women 15-59
Women of working age
Married women
1901
1911
1921
1931
1951
Year
1961
1971
1981
1991
Figure 1. Labour force participation rates for women, 1901-1991
Note: ‘working age’ includes women from the age of 10 in 1901 and 1911, 12 in 1921, 14 in 1931, 15 in 1951,
1961, and 1971, and 16 in 1981 and 1991.
Source: T.J. Hatton, ‘The revolution in women’s work, 1911-1971: a regional perspective’ (Institute of Labour
Research discussion paper 97/08 (1997), Univ. of Essex), tab. 1, p. 27
‘stagnation’,2 less attention has been paid to the measurement of women’s
labour force activity than either before or after. If there was serious
under-enumeration of women’s work in the censuses for this period
then the ‘U’ shape in their participation rates over time would be
correspondingly attenuated.
It has been argued that in the nineteenth century the decisions about
how to enumerate and classify women’s work were greatly influenced by
ideas then prevailing about the proper place of women in the social
order.3 This widely held ideology led to the undercounting of women
who held occupations and engaged in market work or to the wilful
misclassification of them as unoccupied. Horrell and Humphries used
evidence from a sample of household budgets to construct alternative
participation rates for women over the period 1787 to 1865.4 Using
reported occupation alone as the criterion, the results were broadly
consistent with census participation rates (for 1851–71). But when partici-
pation was defined as either having an occupational title or making a
contribution to household income, estimated participation rates were
found to be significantly higher. The more comprehensive definition
was thought more likely to capture intermittent, informal, or low-wage
2 Humphries, ‘Women and paid work’.
3 The so-called ‘cult of domesticity’ for women and the idea of distinct and separate spheres for
men and women lay at the heart of the rise of the male breadwinner model of the family. Such
ideas were widely held among the enumerated as well as by the enumerators—who were overwhelm-
ingly male. For discussions of how these characterizations influenced census enumeration, see Hakim,
‘Census reports’ and Higgs, ‘Women, occupations and work’.
4 Horrell and Humphries, ‘Women’s labour force participation’.
Economic History Society 2001

women’s work in census and survey, 1911-1931
89
employment which would not be recognized in the census as a princi-
pal occupation.5
The main areas of undercounting are often thought to be among
women who worked in family businesses, those in farming families, those
who engaged in part-time work, those who worked as homeworkers or
outworkers, and possibly even among those who held full-time jobs in
factories, workshops, offices, and shops. Higgs made adjustments to the
statistics for the census years from 1851 to 1901, raising the numbers
occupied in agriculture and in all forms of trade and reducing the
numbers of domestic servants, some of whom would not have been
working for wages.6 The effect of these adjustments was to raise the
overall participation rate by 2.2 percentage points in 1851 and by decreas-
ing amounts in subsequent years—thus enhancing the downward trend
in the late nineteenth century.7
Although some sources of under-enumeration were reduced or elimin-
ated by the turn of the century, others may have remained.8 In the
Registrar General’s Office there was a growing awareness of the problem
and in 1911 additional guidance was given on defining women’s occu-
pations. It is worth quoting from the introduction to the 1911 volume
on Occupations and Industries:
The proportions [i.e. participation rates] for females, especially those for
married and widowed, are affected for the purposes of comparison by a new
instruction on the 1911 schedule. In the 1901 schedule occupiers were
requested to ‘State the Occupation, if any, of each person (whether man,
woman or child),’ but this instruction left doubt as to what was required with
regard to the occupations of women assisting their relatives in business or
engaged in domestic duties at home. This gap was filled at the present census
by the following instruction: ‘The occupations of women engaged in any
business or profession, including women regularly engaged in assisting relatives
in trade or business, must be fully stated. No entry should be made in the
case of wives, daughters or other female relatives wholly engaged in domestic
duties at home.’9
5 The information from household budgets appears to cover periods as long as a year and hence
some part of the difference between the two definitions would seem to reflect the distinction between
participation at a point in time and at any time over the year. Anderson, ‘Mid-Victorian census’,
has recently argued that the 1851 census does give an accurate picture of the employment of married
women in the textile trades although he concedes the possibility of under-enumeration of women’s
work in other spheres where employment was casual, seasonal, or part-time. He concludes (p. 27)
that ‘Unless we can find alternative and representative estimates of the proportionate contribution
to family incomes of casual and part-time work, the census enumerators’ books must remain, for
many parts of the country at least, the best indicator that we have of variations in women’s gainful
work activity in the mid-nineteenth century.’
6 Higgs, ‘Women, occupations and work’.
7 In a more detailed study of women in agricultural households, Higgs, ‘Occupational censuses’,
suggested that the numbers of women employed should be increased by a factor of 5 over the
census figures in 1871—although some of these represent transfers from other sectors. He urged
that further studies should be undertaken to see how other sectors might be influenced by a re-
evaluation of the numbers of women employed.
8 Thane, ‘Economic and political status’, p. 97, comments that the 1901 census ‘provides us with
minimum numbers of women in paid work’.
9 Census of England and Wales, 1911, Occupations and Industries, p. cxxviii (italics in the original).
Economic History Society 2001

90
t. j. hatton and r. e. bailey
The report went on to say that the effect of these changes was to
produce some increase in the numbers of women enumerated in agricul-
ture, as boarding- or lodging-house keepers, and as assisting shopkeepers,
dealers, and others engaged in commercial pursuits, but a small decline
in the numbers in domestic service. If all of the increase between 1901
and 1911 (relative to the female population aged 10 and over) in these
sectors is attributed to the change in the instructions for enumeration, it
would account for more than the whole of the increase in the overall
participation rate from 31.6 per cent to 32.5 per cent over the decade.10
It is possible that some further upward adjustments could be made but
it seems unlikely that there would have been severe under-enumeration
in these categories from 1911 onwards.
What about work for wages which, after all, was the bulk of women’s
market work by the turn of the century? It has sometimes been suggested
that, for the reasons outlined above, even women who worked full-time
in factories and workshops on the day of the census failed to appear as
occupied in the census returns. Roberts quotes examples for 1851 to
1881 where, in samples of workers, the majority of women recorded in
the wage books of firms were returned as unoccupied in the census. She
comments that these ‘raise disturbing questions for those using census
data for a study of women’s employment’.11 Higgs gives similar examples
although both authors warn against generalizing from isolated cases.12
Nevertheless, the accumulation of evidence such as this has added weight
to the presumption that there may have been a serious undercount of
work for wages, certainly in the nineteenth century and perhaps well into
the twentieth.13 But a word of caution is called for. Finding women
workers in the wage books of firms can only reveal under-enumeration
since these workers were by definition employed. It cannot reveal cases
where women were recorded in the census as employed but were not, in
fact, in employment.14
10 Ibid., p. cxxxv.
11 Roberts, Women’s work, p. 19.
12 Higgs, ‘Women, occupations and work’.
13 Concerning the period 1871–1931, Roberts, Women’s work, p. 22, observes that: ‘throughout
the period, the combined processes of industrialisation and urbanisation appear to have had little
impact on women’s participation rates although again it must be emphasised that these figures hide
much un-enumerated work’.
14 Other sources which offer the potential for a cross check against the census have not, so far,
proved very decisive. In Walker, ‘Pleasurable homes?’, p. 323, evidence from marriage registers in
south Yorkshire seems to show that these are even more deficient than the census. Walker found
variations among different parsons in the recording, at marriage, of women’s occupations (in
particular domestic service), and August, ‘How separate a sphere?’, p. 288, suggests that the same
might be true of census enumerators in the recording of the occupations of married women.
As Jordan, ‘Female unemployment’, p. 186, notes, industry level aggregate statistics are not likely
to provide an adequate check on the census because of differences in the period of observation and
because of multiple job holding. Differences in industrial classification are also likely to present
problems for accurate comparisons. These limitations are evident in the comparisons made in the
1890s by Clara Collett between data from the census and from the factory inspectors’ returns. She
found that, for those employed in cotton textiles in Lancashire and Cheshire, the census totals for
1891 exceeded those of the factory returns for 1890, and by more for women than for men (Statistics
of employment, p. 35). Similarly, in the woollen trade of Yorkshire and Lancashire the census estimate
exceeded the numbers in the factory return, but this time by more for men than for women (p. 41).
Economic History Society 2001

women’s work in census and survey, 1911-1931
91
Perhaps an even more serious potential source of under-enumeration
is part-time work. To quote Roberts again:
It seems very likely that part-time work (usually undertaken by married
women) was grossly under-enumerated. One reason for this was that part-
time work was so often casual and seasonal (for example of fruit- and hop-
picking which never took place at the same time as the census). It has been
suggested that more part-time jobs might have been enumerated more often
had householders been asked about work rather than about occupation.
Returns were, however, undoubtedly influenced by the widespread assumption
that married women were ‘housewives’.15
Oral evidence suggested that the experience of part-time work was com-
mon among married women in the first decades of the twentieth century.
Roberts found that 40 per cent of respondents in Lancaster, 42 per cent
in Preston, and 50 per cent in Barrow had had part-time jobs at some
time in their married lives. Only 40 per cent in Lancaster, 13 per cent
in Preston and 50 per cent in Barrow had not been engaged in market
work at all since marriage.16 But as the author warns, such evidence
should not be misinterpreted; if these represented only short spells of
employment then at any one time, such as on the day of the census,
only a few would be observed in jobs.17
Part-time work was not separately tabulated in the census until 1951
when it was defined as working for fewer than 30 hours a week. Although
it accounted for only 11.4 per cent of all women workers in the census,
survey evidence suggested that the true figure should be considerably
higher.18 In their study of census participation rates for women between
1951 and 1981, Joshi and Owen concluded that there was systematic
under-enumeration in 1951, 1961, and, to a lesser extent 1981, as
compared with 1971. Using labour force survey data to correct the
figures, they revised upwards the 1951 participation rate for women aged
20–59 (in Great Britain) by 3.8 percentage points. Most of the shortfall
was caused by the under-enumeration of part-time workers—chiefly mar-
ried women.19 The results suggest that the postwar growth in participation
was somewhat slower than the raw census tabulations suggest. This opens
the possibility of similar, or perhaps even greater, under-enumeration in
the censuses of 1931 and earlier when, in contrast to the postwar
censuses, there was no explicit question about part-time employment.
There has been very little discussion of possible biases in census
enumeration in the censuses from 1911 to 1931. While some sources of
undercounting of female workers had clearly been eliminated by 1911 it
15 Roberts, Women’s work, p. 19.
16 Idem, ‘Working wives’, pp. 142–3.
17 It should be noted, though, that if the focus of study is lifetime work profiles, as in Roberts’s
work, rather than the standard labour force concept of those in the labour market at a given point
in time, then such evidence is entirely appropriate. It is important, therefore, to distinguish between
deficiencies in measurement and differences in the definition of what constitutes labour force partici-
pation.
18 Klein, Britain’s married women workers.
19 Joshi and Owen, ‘How long is a piece of elastic?’
Economic History Society 2001

92
t. j. hatton and r. e. bailey
is possible that other sources of bias remained but these have not yet
been investigated. Prior to 1961 the census asked for the individual’s
‘personal occupation’, defined as usually following an occupation or
calling for payment or profit (sometimes known as the ‘gainful worker’
concept). It also asked for details of the current employer, or past
employer if the individual was either out of work or retired. In principle
the unemployed were included in, and the retired were excluded from,
those defined as occupied. But, unlike the later censuses, there was no
requirement to state whether the individual had worked over a specified
period in the recent past and thus the borderline between occupied and
unoccupied is inherently blurred.
Given the looseness of the concept of ‘occupied’ it is legitimate to
suppose that possible biases in the enumeration of women could be
upward or downward. Although there was little change in the nature of
the questions asked, changes in the exact wording, the layout of the
census form, the order of headings, and in the list of instructions might
have influenced the way questions were answered. The appendix summar-
izes the key headings and instructions regarding occupation between 1911
and 1951. It is difficult to infer from these alone what the direction and
magnitude of any bias might be. For example, in 1911 women were
instructed not to be returned as occupied if they were wholly engaged
in domestic duties at home. In later censuses they were instructed not
to be returned as occupied if they were chiefly (1931) or mainly (1921
and 1951) occupied in domestic duties. Thus, in principle, those with
some work were more likely to be returned as occupied in 1911 than in
later years, but it is far from clear how much such distinctions mattered
in practice.20
It is likely that some women with part-time, casual, or seasonal jobs
were recorded as ‘at home’ or engaged in ‘home duties’ even though
they had earned income in the census week or a recent week. But equally,
it is possible that some, without a history of recent employment, could
have been recorded under their former occupation. For such women,
none of the categories ‘out of work’, ‘retired’, or ‘home duties’ may have
seemed entirely appropriate to their occupational status even if one of
the categories might have been a reasonable description of their current
status. Thus, some could have been returned as occupied even if they
would not have been so classified under a labour force survey definition.
This lack of precision in the way in which census questions were posed
does little to sharpen the distinctions in the inherently grey area on the
margins of the labour force—a problem exacerbated by the lack of any
question on when the person was last employed, or (in any census) on
current or recent earnings. This makes comparison with alternative
sources all the more worthwhile.
20 It is possible also that the way in which the questions were asked could have mattered. Thus,
in 1911 those engaged in home duties were not required to make any entry in the columns for
occupation; in 1921 and 1951 ‘home duties’ was to be entered under occupation and in 1931 it
was to be entered in a separate column (see appendix).
Economic History Society 2001

women’s work in census and survey, 1911-1931
93
II
One of the difficulties with evaluating (and correcting) biases in census
participation rates is lack of independent evidence of a sufficiently compre-
hensive nature. For the interwar period there are annual returns for those
insured against unemployment, but the scheme covered only about half
of all women workers and those excluded are more likely to have been
under-enumerated in the census. Thus it is not possible to make the
type of comparisons, and corrections, that Joshi and Owen did for the
census years 1951–81. But one promising source of independent evidence
is the household surveys which, following the basic design of Rowntree
in 1899,21 were undertaken to assess working-class living standards and
to calculate the proportion of households in poverty.
Among the most important of these were the surveys of five towns
(Northampton, Warrington, Reading, Bolton, and Stanley) undertaken in
1912–14 and again in 1924, those for London in 1929–31, Merseyside
in 1929–30, Southampton in 1931, and Sheffield in 1931–2 and, a little
later, those for York in 1936 and Bristol in 1937.22 These surveys were
undertaken by investigators who visited working-class households, usually
according to a predetermined sampling procedure. They recorded details
of the age and sex of household members and relationship to the head,
housing conditions including the number and type of rooms and rent
paid, and the amounts and sources of income. Since the purpose was to
identify poverty, the surveys sought to measure exhaustively the earnings
of all household members, from whatever source and whether part-time
or full-time.
There are advantages and disadvantages to using these surveys to
compare women’s labour market activity with that recorded in the census.
One disadvantage is that they were taken only sporadically, not on census
dates, or even in census years, and they were undertaken for only a few
towns.23 Another disadvantage is that they are not fully comprehensive.
Since the purpose was to measure poverty they covered only the working
class and thereby excluded between one-fifth and one-third of the popu-
lation (depending on the town) who were, by definition, not poor. Given
the exclusion of middle-class households, it might be expected that the
participation rates in the surveys would be higher than in the census.
But against these disadvantages must be set an important advantage: they
provide a unique independent source of evidence, which, like the census,
was based on a household survey, and which recorded information on
labour market status.
Most of the surveys divided the members of each household into
21 Rowntree, Poverty: a study of town life.
22 Bowley and Burnett-Hurst, Livelihood and poverty; Bowley and Hogg, Has poverty diminished?;
Llewellyn Smith, ed., New survey; Caradog Jones, Social survey; Ford, Work and wealth; Owen,
Survey of the standard of living; Rowntree, Poverty and progress; Tout, Standard of living.
23 By definition these surveys exclude agriculture and other rural employments. But agriculture
occupied less that 2% of women by the turn of the century according to the census. On the other
hand, it might be argued that the distinction between participation and non-participation would be
somewhat sharper in urban areas where employment relationships would, in general, be more formal.
Economic History Society 2001

94
t. j. hatton and r. e. bailey
earners and non-earners. Those who normally earned some income but
who were temporarily unemployed, sick, incapacitated, or on strike were
included as earners. Like the census, the surveys excluded as earners
those who were occupied in domestic duties at home but included those
who were assisting relatives in business, who worked on their own
account, or who were working at home for pay or for profit. As an
example, Bowley and Hogg defined earners in the 1924 five-towns survey
in the following way: ‘All members of the household who habitually
worked for wages for at least one full day a week, or who worked for
pay on their own account to the same extent, were described as earners;
other members were non-earners.’24 There seems little to choose, in
principle, between this definition and that used in the census although,
as is shown below, there are some differences in who was counted
in practice.
It might be suggested that the surveys suffered their own biases in
enumeration as a result of poor design or low response rates. In fact this
does not appear to be the case. The surveys were conducted under strict
random sampling criteria and the investigators were instructed to pay
return visits if no one was found at home on the first visit (for example,
if they were at work). In the case of the London survey, only if it was
confirmed that no-one normally lived at the address was another address
substituted.25 Furthermore, the response rates were high even though,
unlike the census, these responses were voluntary: in London the non-
response rate was about 5 per cent and in Sheffield about 3 per cent.26
Thus, the surveys were carefully conducted and there is no reason to
think that there are serious biases in them. Indeed, in two important
respects they might be regarded as superior to the census. First, the
information was collected by an investigator through face-to-face individ-
ual interviews rather than by the respondents filling in a form. And
second, the respondents were often women who presumably would have
given more accurate information on their own working habits than if this
had been reported by a male household member.
One of the most useful surveys for our purposes is that of Bowley and
Hogg which covered five towns in 1924 and compared the results with
those for 1912–14 from the earlier study of Bowley and Burnett-Hurst.27
It reported the labour market status—earner or non-earner—of those
aged 14 and above. The sampling factor for these towns was about one
in nine of working-class households, excluding those in institutions and
24 Bowley and Hogg, Has poverty diminished?, p. 29.
25 See Llewellyn Smith, ed., New survey, III, pp. 31–6, and, for the instructions to investigators,
pp. 413–23.
26 Ibid., p. 32; Owen, Survey of the standard of living, p. 8. It is important to note that the reason
for non-response was generally refusal to give information—not failure to contact the household.
Thus non-response does not imply a bias against sampling households where all members were out
to work. In the London survey it was stated that: ‘The reasons for the refusal of information were
very varied, and (apart from the probable exclusion of the semi-criminal class) there seems little
reason to think that refusals were related to any economic condition, except that there was some
reluctance among the middle class and the better off working class’ (Ibid., pp. 32–3).
27 Bowley and Hogg, Has poverty diminished?
Economic History Society 2001

women’s work in census and survey, 1911-1931
95
Table 1. Participation rates in surveys and censuses
Northampton
Warrington
Reading
Bolton
Women 14 and over
Survey 1912-14
Census 1911
37.9
42.5
27.5
33.1
30.6
34.4
41.4
45.9
Men 14 and over
Survey 1912-14
Census 1911
92.9
93.9
95.5
95.6
93.0
92.7
93.3
93.3
Women 14 and over
Survey 1924
Census 1921
35.4
39.4
30.9
35.2
24.8
31.5
37.4
43.9
Married women
Survey 1924
Census 1911
11.9
11.5
10.0
6.5
6.6
5.4
14.0
14.9
Men 14 and over
Survey 1924
Census 1921
94.2
93.2
94.5
95.6
90.3
89.9
93.2
95.1
Notes: Census figures from 1911 Census, Occupations and Industries, pt. II; 1921 Census, County Tables (Berkshire,
Lancashire, and Northamptonshire). Survey figures derived from Bowley and Hogg, Has poverty diminished?, pp.
72, 94, 120, 148. No comparisons could be made for the fifth town in the survey, Stanley, Co. Durham, because
it is not separately identified in the census.
excluding about 20 per cent of households classified as middle class. For
four of these towns (excluding Stanley in Co. Durham) the survey
participation rates can be compared with those for women aged 14 and
above reported at the nearest census date; either 1911 or 1921.
The figures in table 1 show that the participation rates in the surveys
are lower than in the census, contrary to what might have been expected—
and by a significant margin. For the surveys undertaken in 1912–14 as
compared with the 1911 census, the gap ranges from 3.8 percentage
points in Reading to 5.6 in Warrington. In 1924 as compared with the
1921 census the gaps are slightly larger, ranging from 4.0 for Nor-
thampton to 6.5 percentage points in both Reading and Bolton. Thus,
on these surveys women’s participation rates are about 5 percentage
points lower than those in the census. If the survey results are to
be believed, then the census participation rates would appear to be
considerable overestimates.
It is useful to consider the same comparisons for men and for married
women alone. As table 1 shows, men’s participation rates are closer and
the differences are similar across time although they vary a little across
towns, particularly for Bolton. Thus, whatever the reason for these differ-
ences, they apply much less to men than to women. What about married
women—after all this is the group most often regarded as under-rep-
resented in the census? Unfortunately, married women’s participation
cannot be identified separately in the statistics reported by Bowley and
Hogg, but it is possible to calculate participation rates for married women
and widows under the age of 70. These are reported in table 1 and are
compared with the participation rates for married women from the census.
Economic History Society 2001

96
t. j. hatton and r. e. bailey
These figures are closer but the survey participation rates are likely to
be upward biased—perhaps by about 3 percentage points—as a result of
including widows. If it were possible to make the comparison for married
women alone, the census participation rates would probably exceed those
in the survey. Even so, this would not seem to be the main source of
difference in the overall participation rates.
Table 2. Participation rates in surveys and censuses
London
(1929-31)
Merseyside
(1929-30)
Sheffield
(1931-2)
Southampton
(1931)
Women 14 and over
Survey
Census 1931
33.0
42.8
30.5
35.3
28.0
32.7
23.6
27.6
Men 14 and over
Survey
Census 1931
92.5
92.1
92.5
91.8
93.1
93.8
94.6
90.4
Sources: Census figures from 1931 Census, Occupation Tables. Survey figures from Llewellyn Smith, ed., New survey
(1930-5), vols. III and VI (London, 37 boroughs); Caradog Jones, Social survey, vol. 2, p. 19 (Merseyside, 4
boroughs); Owen, Standard of living, p. 23 (Sheffield); Ford, Work and wealth, p. 110 (Southampton).
It is possible that the difference between the census totals and those
in the surveys could be accounted for by some peculiarity of the Bowley
and Hogg study from which the survey figures are derived. Table 2
therefore compares participation rates derived from surveys undertaken
in the early 1930s with those for the same areas from the 1931 census.28
The area covered by the London survey encompasses 37 boroughs, some
of which were in Middlesex and Essex. The figures for London are
calculated as the weighted averages for these 37 boroughs in both the
survey and the census. The result indicates an excess of 8 percentage
points of the census over the survey, larger than in any of the other
towns. The other towns give discrepancies of about 5 percentage points—
rather similar in magnitude to those found for the four towns in table 1.
Furthermore, the discrepancies in male participation rates are, as before,
relatively small.
These differences between survey and census participation rates for
women seem to be the opposite of what would have been expected from
reading the literature cited earlier. The fact that survey participation rates
are lower than those in the census seems doubly surprising, both because
the surveys counted earnings and because they apply only to the working
class. These discrepancies demand further investigation.
III
On the basis of published tables alone, it is difficult to probe any deeper
into reasons for the striking difference in the measured participation rates.
28 It is not possible to make use of Tout, Standard of living, for Bristol, or Rowntree, Poverty and
progress, for York because participation rates comparable with the census cannot be derived from
the published volumes. In any case these surveys took place at least five years after the 1931 census,
a period in which women’s participation appears to have been increasing.
Economic History Society 2001

women’s work in census and survey, 1911-1931
97
The records need to be examined in more detail. We are fortunate to
be able to do this for one of the surveys: the New survey of London life
and labour (NSLLL). The records from this survive almost in their
entirety (only two boroughs, Tottenham and Walthamstow, are missing)
and they have recently been re-examined and computerized.29 Thus it is
possible to check in detail for possible sources of under-enumeration in
the survey which has the biggest discrepancy with the census. If it is
found that the lower proportion of occupied females in the NSLLL
cannot be accounted for, then suspicion of over-enumeration must fall
on the census statistics. The gap between survey and census can be
explored by posing a series of hypotheses.
Hypothesis 1: Women were systematically misclassified by the NSLLL. On
the original survey cards, some women who were clearly non-earners were
entered initially as earners.30 It seems likely that some were reclassified for
the published results of the NSLLL but this procedure was not docu-
mented. As part of our re-evaluation of the data we have classified women
as in the labour force if they reported an occupation or an employer, or
if they reported earnings last week or in a ‘full week’, or if there were
other reasons (such as unemployment insurance benefits or contributions
or travel to work expenses) to believe that they normally had some wage-
earning work. Those who were assisting relatives in business, who worked
at home, or who had part-time or casual work were also included. The
results are close to the figures for individual boroughs derived from the
published tables of the NSLLL. Overall, we calculate a participation rate
of 33.3 per cent over the 35 boroughs for which we have data as
compared with a participation rate of 43.1 per cent over the same 35
boroughs in the 1931 census. Hence a gap of almost 10 percentage
points remains.
Hypothesis 2: Unemployed women were not counted as occupied in the
NSLLL. The unemployment rate calculated for females in the NSLLL
is 2.9 per cent (of the female labour force) as compared with 6.5 per
cent for the same 35 boroughs in the 1931 census. If the excess were
fully reflected in the labour force and not in differences in employment,
this could account for a difference of 1.25 per cent in participation
rates—about one-eighth of the measured total. But part of the difference
could be accounted for by two factors. First, unemployment was higher
in 1931 than over the period of the NSLLL survey (and hence some of
the higher unemployment in the census would represent lower employ-
29 For details of the computerized records, see Bailey and Leith, ‘New survey’, and Bailey and
Leith, ‘Computerising and coding’.
30 In our analysis of the data, 30.6% of women originally entered on the cards as earners were
reclassified as non-earners. These were often women who were heads of households, or who had
non-wage income of some sort, or who came from a household where there were no other earners.
The most important reason for this is the structure of the survey card, which lists earners first and
then, in the following section, non-earners. Virtually every card has someone listed in the earners’
section. Reclassifying these cases was therefore relatively straightforward and was rarely ambiguous.
Similar misclassification on the original cards seems to have occurred in other surveys: see, for
example, Bowley and Hogg, Has poverty diminished?, p. 29.
Economic History Society 2001

98
t. j. hatton and r. e. bailey
ment and not a larger labour force).31 Second, ‘out of work’ as reported
in the census did not necessarily imply looking for work or seeking
employment as in the definitions used in more recent labour force surveys.
The latter would be closer to the definition implicit in the unemployment
count of the NSLLL.
Alternatively, it might be suggested that the unemployment count in
the NSLLL reflects only those in insured occupations who were receiving
benefit. Women workers in the NSLLL were more likely to be in insured
trades than were women in general—a factor which might have been
expected to increase their unemployment rates compared with the cen-
sus.32 But still it might be suggested that those unemployed but denied
benefit would not be counted in the labour force. In fact, over half those
counted as unemployed in the NSLLL were not receiving benefit and
some of these appear to have been in insured trades.33 It is possible that
even more insured women would have been counted as unemployed at
the date of the census because of the relatively relaxed conditions for
benefit which applied at that time.34 But it also seems likely that the
high rate of unemployment recorded for married women at the census
(20.0 per cent for England and Wales as a whole) includes a significant
share who would not be counted as unemployed on a labour force
survey definition.
Hypothesis 3: Part-time employment was not accounted for in the NSLLL.
The undercounting of part-time workers could have been even more
severe in the NSLLL than in the census. For female earners in the
NSLLL, 86 per cent reported weekly hours (in a ‘full week’) and among
these 7.9 per cent had normal hours of less than 30 per week. A lower
bound, assuming all those with hours not reported were full-timers, would
be 6.8 per cent. This compares with 12.3 per cent of those occupied
31 The interview dates in the NSLLL records indicate that 84.2% of the households were
interviewed before 1931.
32 Occupations excluded from unemployment insurance included agriculture, certain professional
occupations, and domestic service. The under-representation of domestic servants in the NSLLL is
discussed further below.
33 This applies to women aged 16–64. For further analysis of the incidence of unemployment
among both men and women in the NSLLL, see T. J. Hatton and R. E. Bailey, ‘Unemployment
incidence in interwar London’ (Institute for Labour Research discussion paper 98/22 (1998), Univ.
of Essex).
34 In March 1930 the clause requiring that, in order to receive benefit, an applicant must be
‘genuinely seeking work but unable to obtain it’ was abolished. But, under pressure to reduce
expenditure, the government introduced the Anomalies Regulations in October 1931. These were
aimed at reducing successful claims from workers whose attachment to the labour force was thought
to be marginal and they affected married women in particular. Under the Anomalies Regulations
married women who had made fewer than 15 contributions since marriage or, if married for more
than six months, had made fewer than 8 contributions in the past three months, also had to
demonstrate that (a) they were normally employed in insured employment, and (b) they could
reasonably expect to obtain insured employment. These more stringent requirements clearly did
reduce successful claims by married women but they came right at the end of the NSLLL survey
period. Thus the NSLLL and the census were conducted at a time when benefit conditions were
more relaxed than at any other time during the interwar period and hence restrictive benefit
regulations should not be a source of undercounting, at least among women in the insured sector.
For the political and administrative background to the Anomalies Regulations, see Tomlinson
‘Women as “anomalies”’.
Economic History Society 2001

women’s work in census and survey, 1911-1931
99
reporting part-time work (i.e. less than 30 hours) for the same 35
boroughs in the 1951 census. Because no specific question was asked
about part-time work in 1931, those who were chiefly employed in ‘home
duties’ would have been even less likely to report part-time employment
than they were in 1951. Thus, it does not appear that the NSLLL
undercounted part-time employment to a greater degree than the 1931
census. Indeed, it is likely that genuine part-time workers were better
enumerated in the NSLLL than in the census. This would increase the
‘true’ difference between the two sources. However, there were probably
fewer part-time workers in 1931 than there were two decades later and
hence the undercounting of part-timers would be less serious a source of
discrepancy than it might otherwise have been.
Hypothesis 4: Own-account workers were undercounted in the NSLLL. The
1931 census reports the numbers who were ‘working on own account’ and
who were employers or managers but these two categories cannot be
separated at the borough level. It is possible that some census respondents
reported themselves as own-account workers because they had no job
but did not consider themselves unemployed. In the NSLLL it is possible
to identify those who normally obtained income from self-employment.
We classify 4.1 per cent of the female labour force as self-employed and
this compares with 5.8 per cent who were enumerated as either employers
or own-account workers across the same 35 boroughs in the 1931 census.
A significant part of the difference might be attributable to the higher
incidence of proprietors, managers, and employers in the middle class as
compared to the working class.
It could be suggested that both census and survey under-enumerated
own-account workers to some degree. The bulk of such workers recorded
in the census were in the clothing trades (particularly dress and blouse
makers), in retailing, in the professions such as nursing and (music)
teaching, and among keepers of lodging- and boarding-houses, inns or
hotels. Some of them would have been casual workers or outworkers but
qualitative accounts suggest that, due to the effects of mechanization and
Trade Board regulations, outworking and homeworking had declined
substantially since the years before the First World War.35 Both the
NSLLL and the census include (as employed) those assisting relatives in
businesses, even if they were not paid a wage. Other occupations such
as landladies who took in lodgers may not have been as well recorded,
as Davidoff has noted, but these were also in decline by the interwar
period.36 Even if one woman were added to the labour force for every
household in the NSLLL for which income from lodgers is reported it
would raise the participation rate by only 0.6 of a percentage point.
Hypothesis 5: Middle-class women had higher participation rates than
working-class women. Since the surveys cover only women in working-
35 The accounts of homeworking and outworking trades given in the reports of different industries
in the published volumes of the NSLLL illustrate their relative decline (Llewellyn Smith, ed., New
survey, II). See also Pennington and Westover, Hidden workforce, ch. 8.
36 Davidoff, ‘Separation of home and work?’.
Economic History Society 2001

100
t. j. hatton and r. e. bailey
class households, part of the gap could be accounted for by higher
participation rates among middle-class than among working-class women.
It was estimated that, on the definition of working class used in the
NSLLL, about two-thirds of households were working class and the
remainder were middle class. Thus the participation rates of middle-class
women would have to be 30 percentage points higher than those of
working-class women to account fully for the gap in measured partici-
pation rates. While women’s employment in clerical jobs (which
demanded higher levels of education) expanded rapidly between 1901
and 1931, this seems unlikely to account for much of the difference—
indeed, many of these jobs were taken by young women from working-
class households.37 Unfortunately, there is no direct census evidence on
participation rates by social class until 1961. But at that time middle-
class wives were considerably less likely to participate than working-class
wives.38 If the same were true in the interwar period then the omission
of middle-class households from the surveys should have raised the survey
participation rates relative to those in the census, making the anomaly
even greater.
Hypothesis 6: Domestic servants were under-enumerated in the NSLLL.
Much of the difference between census and survey might be accounted
for by the fact that a large share of domestic servants were living in
middle-class households and would not have been enumerated in the
NSLLL.39 Thus, while participation rates may have been lower for
middle-class than for working-class women, the same might not be true
for all women in middle-class households as compared with working-
class households. As a proportion of the female labour force in the 35
boroughs, indoor domestic servants accounted for 19.9 per cent in the
census but only 10.0 per cent in the NSLLL survey. This would imply
that about one-third of domestic servants were non-resident, a figure
broadly consistent with other estimates.40 It does not suggest a significant
underestimate of the numbers which would be expected in working-class
households. If the remaining (resident) domestic servants were added to
the working class it would raise the participation rate of women in the
NSLLL by about 4 percentage points.41 Thus, it could account for about
37 In the chapter on clerical work in the NSLLL it was observed that: ‘the House Sample analysis
has revealed a very large number of clerks living in working class families, and many more of course
are working class in origin. This is the natural result of the Education Acts which, by making
universal a knowledge of the “three R’s”, admitted boys and girls from working class families to at
least the lower grades of clerical employment’ (Llewellyn Smith, ed., New survey, VIII, p. 273).
38 See 1961 census, Household Composition Tables, tab. 42, p. 342.
39 This, of course, means that they would not have appeared in the numerator or in the
denominator of the participation ratio in the NSLLL.
40 A small survey of households containing servants conducted for the NSLLL suggested that
non-resident servants were about one-third of the total in private service and of these about half
were part-time; see Llewellyn Smith, ed., New survey, III, p. 452, and the general discussion of
domestic service in ibid., II, esp. pp. 427–31.
41 If two-thirds of domestic servants were non-resident, then adding them back would increase
the labour force of working-class women by 20%. At the same time they would add 6.7% to the
working-class female population. Thus the participation rate would increase from 0.33 to 0.33*
(1.2/1.067) = 0.37, a difference of 4 percentage points.
Economic History Society 2001

women’s work in census and survey, 1911-1931
101
40 per cent of the difference between the participation rates in the census
and the survey.
Hypothesis 7: Middle-aged women were over-enumerated in the census. As
noted earlier, the early twentieth-century censuses respondents were asked
about their ‘personal occupation’, and not, as later, about whether they
were ‘economically active’. The latter was a rather tighter definition and
explicitly involved having been employed in the census week or, if not
employed, then either intending to get work or sick. It is possible that
in the slack labour market of 1931 (and perhaps earlier) those who were
not retired but who were not currently either employed or unemployed
returned themselves as having an occupation even though they had not
worked recently and were not actively seeking work. This might be
expected to apply especially to women who had been employed at some
time previously. While it might be thought applicable to married women,
the same could apply to single, widowed, or divorced women who had
been employed at some time but who had, in effect, left the labour force.42
Table 3. Participation rates by age in
London
Age
1921 Census,
34 boroughs
NSLLL 1929-31,
34 boroughs
14-15
16-17
18-19
20-4
25-34
35-44
45-54
55-9
45.9
80.6
84.6
70.7
39.9
29.6
28.4
27.1
24.5
20.7
7.3
62.1
89.6
92.8
73.9
28.5
19.3
18.5
19.2
16.4
14.5
5.8
60-4
65-9
70ϩ
Note: The boroughs included are those listed in tab. 4 except
Tottenham and Walthamstow (which are missing from the
NSLLL), and Barking for which age-specific participation is
not reported in the 1921 census.
No breakdown of participation rates by age is available for London
boroughs in the 1931 census, but table 3 compares participation rates
by age groups for the 35 boroughs at the 1921 census with those derived
from the NSLLL for 1929–31. The striking result is that participation
rates among women aged 14–24 are higher in the NSLLL than in the
census, while for women aged 25 and over participation rates in the
NSLLL are lower than in the census.43 The difference in the younger
42 There is no reason to believe that middle-aged women who were working were systematically
under-sampled in the NSLLL. The survey included divided houses and tenements where single,
widowed or divorced women might be living. The only exception would be the resident domestic
servants discussed separately above.
43 In fact the crossover point appears to come at about the age of 30, judging by the participation
rates for England and Wales in 1931 (when the age group 25–34 was sub-divided into 25–9 and
30–4).
Economic History Society 2001

102
t. j. hatton and r. e. bailey
age group can be accounted for largely by a general rise in participation
rates between 1921 and 1931 and by the later labour market entry for
middle-class women.44 But such effects cannot explain why older women
had participation rates substantially higher in the census than in the
NSLLL.45 This difference among older women could explain about 4.5
percentage points of the gap between the census and the NSLLL. And
it may well reflect over-enumeration in the census rather than undercount-
ing in the NSLLL.
IV
One way of further exploring the reasons for the differences in partici-
pation rates is to examine the variations in the participation rate gap
across London boroughs. As table 4 shows, not only did participation
rates vary across the 35 boroughs for which information survives, but the
difference between the census and the survey also varies considerably. In
some cases such as Westminster and St Marylebone the gaps were very
large while in others such as Barking the difference was small.46 It should
Table 4. Female participation rates by London borough, c. 1931
Eastern area
Census
1931
NSLLL
survey
Western area
Census
1931
NSLLL
survey
Bethnal Green
Shoreditch
Stepney
Bermondsey
Hackney
Poplar
Stoke Newington
Deptford
Greenwich
Lewisham
Woolwich
Barking
East Ham
Leyton
Tottenham
Walthamstow
West Ham
48.6
52.0
45.8
44.9
44.5
40.3
43.3
37.6
35.9
33.4
30.8
31.1
33.8
35.3
38.0
36.4
36.8
43.8
44.0
35.2
36.8
37.0
33.7
40.7
30.6
26.5
19.8
18.5
31.0
28.2
27.8
—
Finsbury
Holborn
Westminster
Lambeth
Southwark
Fulham
Hammersmith
Islington
St Pancras
Chelsea
Hampstead
Kensington
Paddington
St Marylebone
Battersea
Camberwell
Wandsworth
Acton
53.0
58.7
58.9
41.8
48.3
41.7
43.0
44.9
48.3
55.0
50.1
52.0
49.4
57.5
37.6
40.7
37.6
40.0
39.1
39.2
46.3
44.5
31.3
31.2
37.6
34.3
36.0
35.1
38.2
38.3
39.6
43.1
34.8
39.2
29.3
32.2
29.5
33.9
28.1
31.4
—
33.3
Hornsey
Willesden
Sources: Authors’ calculations from the records of the NSLLL; 1931 Census, Occupation Tables.
44 For England and Wales as a whole the participation rate of those aged 14–24 rose from 63.1%
to 68.6% between 1921 and 1931. If the same was true for London, then it would account for
two-thirds of the difference between the 1921 census and the NSLLL (5.5 of 8.1 percentage points).
At the higher ages there was no such trend.
45 At the higher ages there was no upward trend between 1921 and 1931: the participation rate
for those aged 25 and over in England and Wales as a whole rose from 23.5% to 23.7%.
46 It should be noted, however, that because the NSLLL samples for individual London boroughs
are relatively small, some of these variations may be accounted for by sampling error.
Economic History Society 2001

women’s work in census and survey, 1911-1931
103
therefore be possible to associate the variation in the participation rate
gap with a series of variables which would help to explain the reasons
behind the observed differences in participation rates.
Table 5. Explaining the participation rate gap (Dependent variable: census
participation rate Ϫ survey participation rate)
Explanatory variable
Coefficient
(t-stat)
Contribution
Constant
Ϫ0.025
(0.41)
Ϫ1.044
(1.93)
0.964
(1.24)
1.816
(2.81)
0.617
(4.70)
0.037
(0.68)
0.091
(0.98)
0.687
2.68
Ϫ0.025
Ϫ0.031
0.011
0.035
0.036
0.007
0.062
Part-time workers/female population 14ϩ (from NSLLL)
Managers and own a/c workers/female popn 14ϩ (in Census)
Ϫ self-employed/female population 14ϩ (in NSLLL)
Unemployed/female population 14ϩ (in Census)
Ϫ unemployed/female population 14ϩ (in NSLLL)
Domestic servants/female population 14ϩ (in Census)
Ϫ domestic servants/female population 14ϩ (in NSLLL)
Share of middle-class households (from NSLLL)
Share of female population 14ϩ aged 30ϩ (in NSLLL)
R2
Heteroscedasticity (␹2(1), 5% = 3.84)
Mean participation rate gap (unweighted)
0.095
0.095
Notes and sources: Dependent variable based on census and authors’ calculations from NSLLL as reported in tab.
4. Part-time workers/female population calculated from NSLLL records assuming that those not reporting hours
were full-timers. Managers and own-account workers in the census calculated from the difference between
‘operatives at work’ and total occupied; in NSLLL from coding of labour force status. Unemployed from census
and from coding of labour force status in NSLLL. Domestic servants in census and NSLLL taken as those
counted under census occupational code 850: Domestic Servants (indoor). Share of middle-class households from
the borough summaries reported by the NSLLL, Llewellyn Smith, ed., New survey, vols. III, pp. 343-412, VI,
pp. 381-462. Share of female population 30ϩ calculated from NSLLL.
Table 5 presents a regression to explain the gap using a number of
variables drawn from the census and the NSLLL. It should be noted
that the object here is to explain differences in measurement rather than
to model the participation decision.47 The dependent variable is the
difference between the census participation rate and our participation
rate calculated from the NSLLL. Overall, the explanatory variables explain
about two-thirds of the variation in the participation rate gap. The second
column in the table is the mean of each variable multiplied by its
coefficient. This shows the contribution of each of the variables to
explaining the overall (unweighted) gap between the census and the
survey across the 35 boroughs used in the analysis.
The proportion of part-time workers to the total female population
aged 14 and over taken from the NSLLL gives a negative sign and is
close to significance at the 5 per cent level. This result indicates that the
larger the share of part-time workers, the lower is the gap between the
47 Hatton and Bailey, ‘Female labour force participation’, provides estimates of behavioural models
of women’s participation using both a cross-section of towns from the 1931 census and a sample
from the NSLLL.
Economic History Society 2001

104
t. j. hatton and r. e. bailey
census participation rate and that in the NSLLL. The coefficient which
is close to 1 implies that part-time workers were not accounted for in
the census. In this respect the census clearly undercounts, even though
this suggests that the true gap is even bigger than the measured gap. In
fact, the overall impact is to widen the gap between the census and the
NSLLL by 3.1 percentage points. The difference between the proportion
of managers, employers, and own-account workers to female population
observed in the census and in the NSLLL also has a coefficient close to
1. This probably represents higher rates of self-employment in the middle
class and it accounts for a little over 1 percentage point of the gap.
The difference in the unemployed to population ratio between the
census and the survey gives a positive coefficient exceeding 1 and it is
significant at the 5 per cent level. This suggests that differences in the
numbers recorded as unemployed represent a difference in the labour
force rather than a difference in measured employment. It seems all the
more likely that this reflects an overcount in the census on a labour force
survey definition. But it might be argued that ‘discouraged workers’ were
undercounted in the NSLLL. Whichever interpretation is adopted, the
effect accounts for 3.5 percentage points of the gap between the census
and the NSLLL.
The difference in the number of domestic servants relative to the
female population observed in the census and in the survey has a positive
and highly significant effect. It accounts for a gap of 3.6 percentage
points between the census and the survey. This clearly reflects the fact
that a large share of domestic servants were living in middle-class house-
holds. But even when this effect is taken into account the coefficient on
the share of middle-class households is positive (although small and
insignificant) rather than negative as might have been expected if middle-
class women had lower participation rates than working-class women.
When the domestic service variable is omitted from the equation the
coefficient on the middle-class variable increases in size and significance.
This adds support to the view that participation rates were higher in
middle-class households largely as a result of the presence of domestic ser-
vants.
The final variable is the share of women aged 30 and over as a
proportion of those aged 14 and over, taken from the NSLLL. It attempts
to capture the difference in reported participation rates among older
women. The cut-off point of 30 was chosen since it appears to be from
around this age that the census and the survey diverge. Although the
coefficient is not significant, it has a potentially large effect, accounting
for 6.2 percentage points of the gap between the census and the survey.
While there is uncertainly about this effect, it does seem to capture a
pervasive difference in enumeration among older women, even when
other possible sources of divergence between census and survey are taken
into account. The constant term which gives a negative coefficient reflects
higher enumeration among younger women in the survey relative to the
census although it too is insignificant.
If we take the more obvious sources of difference between the census
Economic History Society 2001

women’s work in census and survey, 1911-1931
105
and survey figures—unemployed, own account, domestic servants, and
middle class—these together account for gap of 8.9 percentage points.
These might be construed as representing genuine differences in coverage
between the census and the survey rather than differences in enumeration.
Clearly, the negative gap of 3.1 percentage points associated with part-
time work represents under-enumeration in the census. The remaining
gap of 3.7 percentage points is accounted for by the share over age 30
and the constant term. This might be construed as representing genuine
differences in enumeration. The two latter effects roughly balance out,
leaving a small positive excess of 0.6 of 1 percentage point in the census
participation rate over that in the NSLLL. Thus there is no evidence
overall that the census under-enumerated women’s work; in fact, the
opposite may have been the case.
V
There has been a strong presumption in the literature that the census
undercounted a significant share of women’s paid employment. Less
attention has been paid to possible sources of overcounting. The evidence
presented here, based on social surveys, indicates lower participation rates
than does the census for the years 1911, 1921, and 1931. It is worth
reiterating that, since the social surveys attempted exhaustively to docu-
ment sources of household income and to distinguish earners from non-
earners, the results must be taken seriously as an alternative to the census
measure. The purpose of this article is not to offer adjustments to the
census totals but merely to suggest that while there are some downward
biases to the census participation rates there may also be reasons for
upward biases—at least on what seem reasonable definitions of labour
force participation.
The evidence from London boroughs c. 1931 indicates that, to some
extent, these disparate estimates can be reconciled. Differences in cover-
age of own-account workers and domestic servants explain part of the
gap. The results suggest that unemployment was higher in the census
than a strict labour force definition would imply. They also suggest that
the census undercounted part-time workers as many observers have
believed. More tentatively, there seems to have been significant overcount-
ing of employment among women in their middle years whose labour
force attachment had effectively lapsed. On balance, there seems to be
little evidence that the census undercounted women workers as a whole:
in fact it may have slightly overcounted them.
University of Essex
APPENDIX: Questions on occupation in the census, 1911–1951
1911
Separate columns for Personal Occupation, Industry or Service with which worker
is connected, Whether Employer, Worker or Working on Own Account, Whether
Working at Home.
Economic History Society 2001

106
t. j. hatton and r. e. bailey
‘The occupations of women engaged in any business or profession, including
women regularly engaged in assisting relatives in trade or business, must
be fully stated. No entry should be made in the case of wives, daughters, or
other female relatives wholly engaged in domestic duties at home.’
1921
Separate columns for Personal Occupation (k)—including column for education—
Employment (1), Place of Work (m).
‘For a member of a private household (such as householder’s wife) who is
mainly occupied in unpaid domestic duties at home, write “Home Duties” in
Column (k) and leave Columns (1) and (m) blank.’
‘For a member of the household who is chiefly occupied in giving unpaid
help in a business carried on by the head of the household or other relative,
state the occupation in Column (k) as though it were a paid occupation. The
name of the head of the business should be stated in Column (l) as employer,
together with the nature of the business.’
‘For other persons not included above who are mainly dependent upon others’
earnings or upon their own or others’ private means, write “None” or “Not
occupied for a living” in Column (k) and leave Column (l) blank.’
1931
Separate columns for Personal Occupation (K), Employer worked for in occupation
stated in column K, and Employer’s Business (L), For persons (aged 14 or over)
neither usually following an occupation for payment or profit nor retired from any such
occupation, write ‘Private Means’, ‘Home Duties’, ‘School’, ‘Law Student’, etc.,
etc. (M).
‘No entry may be made in these columns [K and L] for wives or other
relatives chiefly employed in unpaid domestic duties at home, . . . or persons
dependent upon their own or others’ private means. For persons in these classes
an entry should be made in column M instead. . . A member of a household
who is chiefly occupied in giving unpaid help in a business carried on by the
head of the household or other relative should state the occupation in column
L, treating the head of the business as employer.’
1951
Separate columns for Personal Occupation (P), Employer and Employer’s Business
(R), Place of Work (S).
‘Unpaid Helpers in a Family Business. For a member of a household who
is chiefly occupied in giving unpaid help (i.e. without receiving a definite wage
or share of the profits) in a business carried on by the head of household or
other relative state the occupation on Column P, adding “Unpaid”, and Com-
plete columns R and S treating the head of the business as employer. Do not
include persons whose help is given for private domestic duties only.’
‘Home duties. For anyone chiefly occupied in unpaid domestic duties at
home write “Home duties” in Column P and leave Columns R and S blank.’
Footnote references
Anderson, M., ‘What can the mid-Victorian census tell us about variations in married women’s
employment?’, Loc. Pop. Stud., 62 (1999), pp. 9–30.
August, A., ‘How separate a sphere? Poor women and paid work in late-Victorian London’, J.
Family Hist., 19 (1994), pp. 285–309.
Bailey, R. E. and Leith, A., ‘New survey of London life and labour: codebook for the datafiles’,
(Release 1.0), (1997), University of Essex.
Bailey, R. E. and Leith, A., ‘Computerising and coding the New survey of London life and labour’,
(1997), University of Essex.
Economic History Society 2001

women’s work in census and survey, 1911-1931
107
Bowley, A. L. and Burnett-Hurst, A. R., Livelihood and poverty (1915).
Bowley, A. L. and Hogg, M. H., Has poverty diminished? (1925).
Davidoff, L., ‘The separation of home and work? Landladies and lodgers in nineteenth and twentieth
century England’, in S. Burman, ed., Fit work for women (1979), pp. 64–97.
Ford, P., Work and wealth in a modern port: an economic survey of Southampton (1934).
Hakim, C., ‘Census reports as documentary evidence: the census commentaries, 1801–1951’, Socio-
logical Rev., 28 (1980), pp. 551–80.
Hatton, T. J. and Bailey, R. E., ‘Female labour force participation in interwar Britain’, Oxf. Econ.
Pap., 40 (1988), pp. 695–718.
Higgs, E., ‘Women, occupations and work in the nineteenth century censuses’, Hist. Workshop J.,
23 (1987), pp. 59–80.
Higgs, E., ‘Occupational censuses and the agricultural workforce in Victorian England and Wales’,
Econ. Hist. Rev., XLVIII (1995), pp. 700–16.
Horrell, S. and Humphries, J., ‘Women’s labour force participation and the transition to the male
breadwinner family, 1790–1865’, Econ. Hist. Rev., XLVIII (1995), pp. 89–117.
Humphries, J., ‘Women and paid work’, in J. Purvis, ed., Women’s history: Britain, 1850–1945 (1995).
Jones, D. Caradog, The social survey of Merseyside (3 vols., 1934).
Joshi, H. and Owen, S., ‘How long is a piece of elastic? The measurement of female activity rates
in British censuses, 1951–1981’, Camb. J. Econ., 11 (1987), pp. 55–74.
Jordan, E., ‘Female unemployment in England and Wales, 1851–1911: an examination of the census
figures for 15–19 year olds’, Soc. Hist., 13 (1988), pp. 175–90.
Klein, V., Britain’s married women workers (1965).
Smith, H. Llewellyn, ed., The new survey of London life and labour, 9 vols. (1930–5).
Owen, A. D. K., A survey of the standard of living in Sheffield (Sheffield, 1933).
Pennington, S. and Westover, B., A hidden workforce: homeworkers in England, 1850–1985 (1989).
Roberts, E., ‘Working wives and their families’, in T. Barker and M. Drake, eds., Population and
society in Britain (1982), pp. 140–71.
Roberts, E., Women’s work, 1840–1940 (1988).
Rowntree, B. S., Poverty: a study of town life (1901).
Rowntree, B. S., Poverty and progress (1941).
Thane, P., ‘The economic and political status of women’, in P. Johnson, ed., Twentieth century
Britain: economic, social and cultural change (1994).
Tomlinson, J., ‘Women as “anomalies”: the Anomalies Regulations of 1931, their background and
implications’, Public Admin., 62 (1984), pp. 423–37.
Tout, H., The standard of living in Bristol (Bristol, 1938).
Walker, A., ‘Pleasurable homes? Victorian model miners’ wives and the family wage in a south
Yorkshire colliery district’, Women’s Hist. Rev., 6 (1997), pp. 317–36.
Official publications
Board of Trade (Labour Department), Report by Miss Collett on the statistics of employment of women
and girls (P.P. 1894, LXXXI).
Census of England and Wales, 1911, Occupations and Industries, pt. II (P.P. 1913, LXXIX).
Census of England and Wales, 1921, County Tables (1924).
Census of England and Wales, 1931, Occupation Tables (1934).
Census of England and Wales, 1931, County Series (1932–3).
Census of England and Wales, 1961, Household Composition Tables (1966).
Economic History Society 2001