Women , Occupations and Work in the Victorian Censuses Revisited by Edward Higgs and

There has long been a tendency amongst historians to view the Victorian and Edwardian censuses of England and Wales as a problematic source for studying the work of women. Census-taking in the period was a predominantly male affair – census enumerators, who were mainly men, gave to household heads, again mostly male, census household schedules which they filled up using instructions provided by the exclusively male civil servants of the General Register Office (GRO) in London. The Victorian enumerators collected the household schedules and copied them into census enumeration books (CEBs), and then dispatched these to the officials at the GRO. When the latter received the CEBs they proceeded to ‘abstract’ the information in them using classification and coding systems they had devised to create tables and commentaries to be published in Parliamentary Papers. This, it has been argued, introduced biases against recording the work of women at almost every stage. If such under-enumeration existed it would create signal problems for understanding the changing role of women in the economy and in the family, and indeed the nature of economic development during the Industrial Revolution as a whole. This article is in two parts. In the first, Edward Higgs examines the historiography on the issue, and his own position in it, and in the second, Amanda Wilkinson presents new evidence on the reliability of the census returns.

considered heads of households'. 3 Similarly, in History Workshop Journal 35 (1993), in 'Women, Work and the Census: a Problem for Historians of Women', Bridget Hill, while claiming 'censuses are -or should be -a way in to knowledge of work done by women', asserted that, 'what was common to all censuses was that women's work was consistently under-recorded'. 4 In 1995, Sara Horrell and Jane Humphries argued in 'Women's Labour Force Participation and the Transition to the Male Breadwinner Family, 1790-1865', that although 'many authors have used nineteenth-century census data to demonstrate declining female participation and increasing employment segregation . . . the census enumeration of women's employment is demonstrably inaccurate'. According to Horrell and Humphries, the most substantial underreporting was to be found in the case of married women, in the agricultural sector, in manufacturing and in certain service occupations. Such work, they concluded, was 'invisible' to male observers. 5 Instead they based their analysis of women's participation in the labour force on family budgets in the works of contemporary social commentators, Parliamentary Papers, working-class autobiographies, and similar sources. 6 Such a strategy is perhaps understandable given the fact that they were primarily concerned to examine the period when censuses did not give information for individual occupations. However, this did not stop them from querying the usefulness of censuses in general. The same year Humphries, in her contribution to June Purvis's Women's History, repeated these claims regarding the problems with the census, and showed that her and Horrell's budgets recorded far higher levels of labour participation for women than in the census tables. 7 However, their figures on women with occupational titles from the same sources showed similarities with my revised census figures for women's employment in 1841. 8 More recently Alison Kay rejected using the census for the study of women retailers in nineteenth-century London on the grounds that the source 'suffers from a number of well-documented flaws', and turned instead to the use of insurance records. 9 The same year, 2006, Hannah Barker in her examination of late eighteenth and early nineteenth century female enterprise in Northern England also claimed that it 'has been well established that the census is particularly problematic in the case of women's work'. 10 Returning to the subject of the usefulness of the census in an article with Carmen Sarasu´a in Feminist Economics in 2012, Jane Humphries again claimed that the work of women was 'left off the record' in censuses across the world, including those in Britain. From this she and Sarasu´a argue that the apparent U-shaped curve in women's participation rates in the economy over the period of industrialization is a simplistic rendering of the reality of women's lives, and in part a 'statistical artefact' of official tables. 11 It is clear, therefore, that there is a considerable body of opinion that holds that the censuses under-enumerated the work of women during the Victorian era. What is less clear, however, is where the factual evidence for this conclusion with respect to the British censuses, at least in terms of occupational titles, is to be found. How has it been shown that they are 'demonstrably inaccurate'? Hill and Rose appear not to have carried out any research on the source itself, Hill basing her comments on the work, primarily, of Davidoff and Hall, and a 1987 article of mine in History Workshop Journal, 'Women, Occupations and Work in the Nineteenth Century Censuses'. 12 Although earlier research had raised issues about particular aspects of the problem, my article was one of the first works to confront directly the problems of women's work in the nineteenth-century British censuses. Horrell and Humphries, and Humphries and Sarasu´a, also cite my work, whereas Davidoff and Hall do not back up their general statements. Kay and Barker also cite me as the source for their comments on the Victorian censuses. In summary, it seems that the assumption that the work of Victorian women in the British censuses is under enumerated relies to a worrying extent on the comments made by me some thirty years ago, which I subsequently repeated in my guide to the census records, Making Sense of the Census in 1989. 13 There have been a number of attempts since then to rehabilitate the Victorian census as a source for women's work as a riposte, in part, to my original arguments, and to those of historians who agreed with me. In 1998 John McKay noted that the published Census Reports indicated that there was consistently higher employment for married women in nineteenthcentury Lancashire than elsewhere, and in industrial areas within that county rather than in rural area, all of which pointed to the usefulness of the source. 14 In 1999 Michael Anderson showed from research based on samples of the CEBs for Lancashire in 1851 that large numbers of women continued to work in the textile factories after they had married. They only appeared to move out of such employment when they had children and had no-one in the household to provide childcare. His conclusion was that the manuscript censuses were a useful source for women's work. 15 This subsequently led Leigh Shaw-Taylor to argue that the married work of women in factories was very well recorded, and that it was unlikely that unmarried women's work would be less well enumerated. 16 However, it is perhaps unwise to base statements about all women's work on the basis of Lancashire alone, because, as Shaw-Taylor himself stresses, the experience of women was very diverse, and 'any simple aggregate national narrative about women's experience of the labour market during the industrial revolution is likely to be seriously misleading'. 17 As Nigel Goose has recently summarized the situation, 'the more informed critics of the census data have concluded that the jury is out on the question of under-enumeration [of women's employment]'. 18 Given the use that has been made of my early work, and the controversy it has created, it would perhaps be useful to revisit my original arguments, especially in the light of a more recent attempt to gauge the reliability of the census data. 19

REVISITING 'WOMEN, OCCUPATIONS AND WORK IN THE
NINETEENTH CENTURY CENSUSES' In my 1987 HWJ article I identified what I regarded as a number of shortcomings in the recording of the work of women in the Victorian censuses. They included the problems of casual and seasonal employments; the difficulty of determining whether women's work in the home was part of the market economy or not; and the influence of the Victorian ideology of separate spheres for men and women on the exclusion of women from the recorded occupied population. As already noted, Victorian census-taking was a predominantly male activity, and this, I argued, is crucial for understanding how the economic activities of women were recorded, or not, as the case might be. I covered a lot of ground, bringing a good deal of evidence to bear on my subject, but my overall conclusion was not in fact a direct rejection of the usefulness of the census records, while some of my arguments were perhaps, with hindsight, not as grounded as they might have been.
First, I pointed out that the census schedules given to Victorian householders to fill out indicated that the work of women was only to be recorded as an occupation if they were 'regularly employed from home', thus leading to the omission of much casual work from the census schedules. 20 I also gave considerable emphasis to the fact that the Victorian censuses were usually taken in March or April to avoid the movements associated with the arable harvest. This, I argued, must have led to the under-enumeration of seasonal labour performed by women. Also, since the published census tables could only deal in single occupations, the multiple activities of women were underestimated. 21 My broader point was that the census recorded occupational titles, which were as much social designations as the modern concept of 'gainful employment' -occupational titles recorded what people were called or designated as much as what they did. 22 Although there is much to these arguments, it should however be noted that such problems must have applied to some extent to men and boys as well as to women and girls. In a somewhat less well-cited article, I pointed out that large numbers of men and whole families were active in seasonal agriculture in the Victorian period, whether in the cereal harvest or in hoppicking. Whatever their occupations in the CEBs, they would not be likely to figure as 'seasonal hop pickers', and the term does not appear in the published Census Reports. 23 My conclusion was that although labour inputs by men as well as women into agriculture in the Victorian period are probably underestimated by the census because of the omission of seasonal and casual labour, any likely revisions would not materially alter the picture of England as having, in global terms, a uniquely small agricultural sector in the mid nineteenth century. 24 Moreover, as Raphael Samuel noted many years ago, the economic activities of many Victorian men were often as irregular and seasonal as those of women and children, and we should not base a picture of their economic situation solely on the experience of factory production, especially in the North of England. 25 Nor should we over-estimate how geographically widespread the large-scale factory production was, at least in the early part of the nineteenth century. In the 1831 census, parish clerks were asked how many people were occupied (1) in agriculture; (2) in manufacture for 'export' out of the locality; (3) in retail trade or handicrafts for sale in the locality; (4) as capitalists, merchants and professionals; (5) as miners, fishermen, non-agricultural labourers; (6) those retired or disabled; (7) and finally as servants. Tony Wrigley's examination of the published results of this enumeration shows that nearly two thirds of the manufacturing population in England were to be found in as few as twenty-four out of the nearly 700 hundreds (county subdivisions), or their equivalents, in the country as a whole. 26 Perhaps when conceptualizing occupational structures in the period we ought to think as much about the worlds of William Cobbett's Rural Rides and Henry Mayhew's London Labour and the London Poor, as about that of Friedrich Engels's Manchester. Calculating such labour inputs is certainly a problem, but despite this no-one has ever suggested abandoning the use of the census to examine the work of men, and such problems are still a feature of census-taking today. Amanda Wilkinson's findings, discussed below, also indicate that such deficiencies were not as severe in the original census enumerators' books as one might fear. However, a detailed revision of the figures to provide a total for the number of all persons economically active in Britain from the nineteenthcentury censuses is plainly beyond the scope of an article such as this.
Implicit in some of what I and others argued about the underenumeration of women in the Victorian censuses was the belief that how male householders, census enumerators and the census authorities treated women's work was influenced by the Victorian view that woman should be the domestic 'angel in the house' rather than entering the external, masculine world of work. Essentially, or so the argument went, the fact of women working outside the home was seen as shameful, and so what work they did would be hidden, or would not be seen as an occupation in census terms. The classic discussion of this ideology of 'separate spheres' for the nineteenth-century middle classes is, of course, the 1987 work of Davidoff and Hall, but how far could similar attitudes be found in the working classes at this time? Davidoff had herself, in a contribution to a collection in 1979, queried the Victorian separation of home and work in the case of landladies. 27 Much recent work has also undermined the concept of a strict separation between the work of men and women, at least among middle-class women, offering empirical evidence on women's business roles in nineteenth-century Britain. Here we might include the collection Women, Business and Finance in Nineteenth Century Europe: Rethinking Separate Spheres, edited by Robert Beachy, Beatrice Craig and Alastair Owens; Hannah Barker's The Business of Women; Alison Kay's The Foundations of Female Entrepreneurship; Women in Business, 1700-1850, by Nicola Phillips; and the work of Jennifer Aston. 28 Some of these authors however, as already noted, draw directly upon my own work to point to the supposed problematic nature of the census as a source. In the London working class Andrew August, through an analysis of late Victorian census enumerators' books from the East End, showed already in 1994 that 'many married women' rather than embracing the 'separate spheres ideology, and the male breadwinner ideal', had 'earned wages when their household economies were relatively stable'. Nevertheless, he too noted apparent differences in how far census enumerators were prepared to include the work of women in the returns, and claimed that this made his figures for women's labour-force participation a low estimate. 29 Much of what I said in my 1987 article about the influence of separate spheres ideology on women's work in the census related to the occupational tables in the published Census Reports, rather than to the data in the original household schedules and CEBs. I noted that between 1851 and 1871 women working at home on 'domestic duties' were to be included in the tables in the Census Reports showing 'the occupations of the people'. In the same period the wives of innkeepers, lodging-house keepers, shopkeepers, butchers, farmers and shoemakers, were included under special terms (with 'wife' added to the husband's occupation as in 'shoemaker's wife') in the appropriate section of the tables because they were 'supposed to take part immediately in their husband's business'. However, such categories disappeared from the published tables from 1881 onwards, and women working in the home were also placed in a new 'Unoccupied' class. 30 I described too how Charles Booth, when in the 1880s he attempted to create a consistent series of occupational totals for the period 1801 to 1881, simply pushed this exclusion of women back into the figures before 1881, a strategy accepted by many subsequent historians. 31 I did look at some original CEBs, from Colyton in Devon, Spitalfields in London, and Matlock in Derbyshire in 1851 and 1881, and noted variations between differing enumeration districts with regard to the occupations of women, and their proportions in work. The variations, I suggested, might 'again reflect a particular enumerator's habit of ignoring the paid work of women rather than a low economic participation rate'. 32 However, I did not address any other possible reasons for the geographically 'spotty' nature of women's employment, such as differential proximity to work places, or work being put out to family or neighbours. Perhaps I was guilty of applying the ideas of Davidoff and Hall on separate spheres without attention to their detailed approach to archival research, and to context and specificity.
Moreover, some of my criticisms of the recording of women's work in the census returns related not to under-enumeration but to mis-enumeration. Drawing upon research for my doctoral thesis, I pointed to the numbers of 'housekeepers' in Rochdale and Rutland in 1871 who were resident in the homes of kin, or even the heads of householders, and were probably not domestic servants but working in the family home. I also drew attention to the number of 'domestic' servants working in the households of retailers and farmers, who may have been active, at least part-time, in retailing or agricultural tasks. 33 However, these may again be problems associated primarily with the published tables, and might easily be adjusted for when considering the original manuscript returns. Even this may have been taking the argument too far since, as Michael Anderson subsequently showed, the clerks at the Census Office in London seem to have been aware of such problems and made allowance when abstracting data for the published Census Reports. 34 I brought together much of the evidence that was available at the time from those who had used CEBs to reveal problems with women's census occupations in the Victorian period -my own detailed work on domestic servants in Rochdale; 35 Judy Lown's work on the silk industry in Essex; 36 Walton and McGloin's study of landladies in Keswick in 1901; 37 Jessica Gerrard's study of the casual employment of villagers in country houses; 38 Miller's study of field workers in Victorian Gloucestershire; 39 and John Holley's study of married women in two businesses in south-east Scotland. 40 However, this coverage was somewhat episodic and patchyhow typical were the problems revealed, often in passing, by these local studies? After all, I did not list all the local studies that did not meet similar problems and thus had not raised such methodological issues. Given the resource constraints in the 1970s and 1980s associated with inputting and analysing the data from the CEBs on computer, any examination of such issues had to be at the local level, and was thus inherently prone to such limitations. Amanda Wilkinson's evidence regarding women's employment in the British census based on case studies, given below, covers a wider geographical area, and approaches the issue with new techniques of analysis.
However, despite such problems, in my 1987 article I counselled using the census returns with care, rather than steering clear of the source a together: The conclusion to be drawn from this work is that it is necessary to treat the occupational information in the manuscript census enumerators' books with caution, and that the historian's use of the published census reports should be even more circumspect. Without a knowledge of local economic and social conditions, to give one a feel for possible problems of underenumeration, and a grasp of the shifting administrative conventions of census tabulation, the use of these sources may be fraught with dangers. 41 I expressed similar reservations in Making Sense of the Census two years later. However, by 2005, when a revised and updated version of this work was issued as Making Sense of the Census Revisited, although still cautious I struck a more positive note, claiming that 'in the absence of alternative sources, the census enumerators' books are still our best source for understanding the economic activities of women in the Victorian period'. 42 As the work of Humphries, Horrell and others have shown, other sources do exist for studying this subject, so the question is really whether the census is a good enough source to be used profitably.
The next section of this article will present evidence from the work of Amanda Wilkinson which suggests that my increasing confidence in the CEBs as a source for the history of women's work perhaps still did not go far enough. Given that the problems with women's employment in the Census Reports have been well chronicled already, and that the British CEBs for the period 1851 to 1911 are now available in a single machine-readable dataset, 43 the published census tables are less vital for historical research. But it is therefore important to establish the reliability of the CEBs above all else. 44 Here new evidence on the reliability of the census returns in East Anglia and London will be presented first, and problems with examining women's work via small-scale studies will also be addressed.
We hope that this will help to tip the balance so that Nigel Goose's hypothetical jury of historians judging the reliability of the census will come down provisionally on the side of 'Not Guilty'.

CASE STUDIES
Admission records relating to county asylums are a source rarely used or consulted for this purpose. Much of the detailed research carried out to date on the recording of women's work in nineteenth-century England and Wales has been small in scale and very local in nature, focusing on the records of a single employer, or on the census enumeration books relating to a small community. This has created difficulties when small-scale research studies are then used to generalize about the enumeration of women's work. A regional perspective, as in the following analysis for East Anglia and London, may help to overcome some of these problems. But how on this scale does one find other sources to corroborate the evidence in the CEBs? Despite extensive searches through the wage books and employment records held in regional record offices across the east of England and London, very few suitable sources have been found. For this type of record linkage with the censuses each source needs to supply for every female employee her name, an address, and if possible her age, to ensure that cross referencing with the CEBs and other sources is reliable. Most of the employment records available do not contain full names: the majority give initials, or even just the number of people working on a given day. Very few of those which do provide names will give addresses, or any other means of categorically identifying individual workers. Given, therefore, the limited number of sources against which the CEBs can be cross referenced, it has been necessary here to turn to other sources that give information on the work carried out by women at the time of a census.
The asylum records examined for the present study covering the main provincial towns of East Anglia -Norwich, Colchester, Brentwood and Ipswich -contained information on every woman admitted to the County Asylums. These were St Andrew's Hospital near Norwich, St Audry's in Melton near Ipswich, and Warley Hospital near Brentwood in Essex. By searching the entire asylum population for all women admitted to the asylums within the three months following a census, it was possible to ascertain the names of these women, their ages, their occupations prior to admission, secondary information on their immediate families, and the address at which they were in residence before entering the asylums. 45 We assume that women who were given occupational titles were employed in the appropriate work: it is difficult to see why they would have lied since there is no suggestion that women would get better treatment if they worked. The 'three month rule' was applied to ensure that, as far as possible, the statistics would not be affected when women changed their economic status. It was then possible to use this information to trace many of these women in the relevant censuses, and to compare how well the occupational details in the registers matched those in the CEBs.
An entry was considered a match if the details of occupation given in the patient register corresponded exactly with those in the CEBs, and ideally both in terms of address and kin. The matches shown in Graph 1 relate to women who were recorded as being married or widowed on admittance to the asylums, and include both those who were recorded as having no occupation, and those who were employed. In some instances, where it was possible to identify a woman through her kin, then these were also classified as a match. Furthermore, in a small number of cases included in the figures under matches what could be regarded as a partial match sometimes occurred. Martha Miller, for example, was recorded in the patient records as being a 'Blacksmith's wife' (therefore unemployed), but in the CEBs she was recorded as being 'formerly a dressmaker'. She was still unemployed but more detail was given in the CEBs in this instance. Whilst this could be described as a match of employment status rather than of occupation, this in itself is still a valuable result, since it is important to match those who were not working as well as those who were, if a full picture of employment status is to be obtained through the study of the censuses. Of course, these women were relatively poor, and we would not expect middle-class women to be represented, but if the women were recorded with occupations then they were plainly not affected by separate-spheres ideology.
It can be seen clearly in Graph 1 that the percentage of instances where information given in the patient registers matched that given in the censuses was high, and increased over time. In Ipswich in 1881, for example, 100 percent of the details given relating to married and widowed women matched.
In each asylum servants -housekeepers, housemaids, and the like -made up a disproportionately high number of recorded admissions. After these, dressmakers, charwomen, laundresses, washerwomen and school mistresses seem to have been the most likely candidates for suffering mental illness. While silk weavers, stay-makers, tailoresses and other 'key' occupations were also admitted, their numbers were very small in proportion to those present in the towns as a whole. Many of the unoccupied married women were recorded as being 'wife of labourer', or 'wife of agricultural labourer', indicating their low social standing. The preponderance of lower workingclass occupations is, of course, to be expected, since the public County Asylums were intended for those who were unable to afford medical treatment. We are looking primarily at the very poorest women in society, and perhaps unsurprisingly these were the women who were suffering in the majority of cases with 'exhaustion of melancholia' (what we would probably refer to now as severe, chronic depression).
The probable reason for the high match rate between the two sets of sources, is explained by some of the Warley records for 1871. The case books for the early years of the Warley Hospital (prior to c.1875) no longer exist, and all that remain are the 'Notice of Admission' forms that were used in the pauper asylum system to section the patients. 47 These detail the physician's findings and the orders from the Justice of the Peace for the patients to be admitted. The personal details were given by the head of the household in which the patient was staying, normally the husband, brother or mother. The person giving the woman's details was, therefore in many cases, the same person who completed the census schedule perhaps only days before. The details in the asylum registers were taken down not by an enumerator, but by a Justice of the Peace and a medical officer. It seems somewhat improbable that a next of kin trying to deal with the sectioning of a loved one in an acute state of ill health, and most likely frightened and distressed themselves, would supply misleading information to save face, or to fit in with an ideology of the middle classes. Of those admitted (including married, widowed and single women), 39.9 percent of women in Norwich are recorded as having an occupation, 42.9 percent of the Ipswich women, and 27.3 percent of those in Essex. 48 Although the ill health of many of the poor women admitted would have affected their ability to work, it might be expected, therefore, that the numbers showing an occupation would be relatively high. In the enumeration districts examined, on the other hand, there will have been some women who were in the fortunate position of being able to choose not to work, and, as such, it might be expected that the numbers shown as working in these districts might be lowered as a result. The fact that the figures are as high as they are makes it improbable that there were large numbers of women whose formal work had gone unrecorded by the census.
It might be argued that all this is a study in language, in what people were called, and that such studies must always be trumped by those that compare census data to records that deal with actual employment, wherever they exist. However, because of the paucity of good employment records from the period (already noted), any studies based upon them must necessarily risk being geographically limited. An example of these problems can be seen in Lown's study of women working in the Courtauld silk mill in mid nineteenth-century Halstead, Essex. Lown consulted Courtauld's employment registers and obtained a random sample of 1,009 workers, taken from a total of 172 households, who were working in the mill in 1861. Out of these workers, 575 were female, most apparently single. One hundred and sixty of these households were located in the census returns for Halstead, another two in the village of Little Maplestead, one in Great Maplestead, one in Castle Hedingham, one in Colne Engaine, and one in Stisted (leaving five households unaccounted for). In the course of her research, Lown attempted to cross reference the details of the workers she had found in the Courtauld employment registers with the census returns of 1861. She concluded that, of the married women working in the mill, 'not a great many evade [occupational] classification', 49 since the majority of the women she traced were recorded as having the same occupation in the census returns and the employment records.
It is perhaps surprising, therefore, that she also claimed that it was the norm for enumerator bias to play a part in occupational enumeration, and that women's occupations were normally left blank, when her own findings suggested the opposite. This may be explained by her observation that 'straw plaiting evades classification'. 50 Straw plaiting is known to have been a primary occupation for women living in the Halstead area of Essex in the latter half of the nineteenth century, and its omission from the census returns would certainly prove that the occupational enumeration of married and single women was problematic, if not hopelessly compromised. Lown suggested that the lack of straw plaiters in the census was likely to have been due to the fact that straw plaiting, being an occupation where much of the work was carried out at home, was not considered 'regular' work. Therefore, she suggested that the women concerned were not considered to be fully engaged in employment, and that their occupations thus did not need to be reported in the census returns. This aside has been used by some later historians to cast doubt on the reliability of the census returns as a whole.
Lown's argument would initially appear well founded, but her findings were misleading due to the problems related to using a small sample in a restricted locality. Her sample was located primarily in Halstead, with the women traced in the census returns living in the environs of the silk mill in the centre of a fairly large market town. As noted above, only six of her sample households were in the villages around the town, and these were families who were linked to the mill -mill workers, not straw plaiters who tended to be the wives and children of agricultural and rural labourers. Thus, it was not that straw plaiting was being ignored by the enumerators as she suggested, rather it was that Lown was looking in the wrong place for them. Research on the villages around Halstead shows that straw plaiters most certainly were present, and in great numbers. In each of the years 1851, 1861, 1871 and in some cases 1881, more women were recorded as straw plaiters in the villages around Halstead than in any other female occupation. In the village of Castle Hedingham in 1861, the year on which Lown's research was based,163 female straw plaiters were recorded in the census returns: thirty-five percent of women over the age of fourteen. The next most populous occupation title was, unsurprisingly, 'domestic servant' with twenty-seven women so recorded. 51 Likewise in Little Maplestead 52 and Great Maplestead straw plaiting was by far the largest reported occupation for women in the censuses. 53 In a cluster of four villages studied in the Halstead area the only one which did not follow the trend was Pebmarsh. This community was centred around a silk throwing mill, with ninety-two women recorded as silk throwsters, and, as might be expected, only twenty straw plaiters. 54 Using a small sample from a single employer, or census manuscripts or reports from a single community, is thus fraught with problems, and cannot reliably represent a single area, let alone an entire nation over time. It is also of interest here that the censuses were indeed recording work that was casual and based on the home. * * * The levels of poverty and the bad living conditions experienced by many of the London working classes during the latter half of the nineteenth century were such that for large numbers of families, the 'luxury' of the wife simply being a housewife and mother was not a possibility. The image of the angel in the house, the domestic goddess of Victorian domestic ideology, was something that those living in the slums of the East End could only dream of, if it occurred to them at all. It may not have been considered 'correct' for women to work during the Victorian age, but the 'lady of leisure' was a middle-class ideal, not a working-class reality.
There is little doubt that a large proportion of women did carry out paid work in order to survive. Ellen Ross noted that A. L. Bowley's pre-First World War study, which examined working households in twelve British towns, found that, 'only about five percent of unskilled workers' households could survive on the man's wages alone'. 55 This left something in excess of ninety percent of such households requiring a supplementary wage in order to escape the workhouse or starvation. It would be wrong to suggest, however, that ninety percent of women would therefore have to be working, as a large number of these households would have been home to children of working age who could themselves supplement the income of the breadwinner. In Family Structure in Nineteenth-Century Lancashire, Michael Anderson described how in households working in mills in Preston, when the parent was over the age of thirty-five, sixty-nine percent had at least one co-residing child in employment. 56 Ross argued in 1993 that 'the large married women's work force in London, often unlisted by census enumerators either because the male ''household head'' failed to mention it, or because the census taker viewed the wife's work as insignificant, has remained largely invisible even today'. 57 However, it is not at all clear that this is the case. A study of all women in four enumeration districts (respectively in Bethnal Green, Spitalfields, Camberwell, and Saffron Hill) suggests that their employment was not invisible, nor un-recorded in the CEBs. The findings summarized in Graphs 2-5 suggest that the census offers a window into women's work in London during the nineteenth century, and accurately tracks the changes in occupational opportunities that women experienced.
A comparison of the graphs reveals marked differences between these districts, and also distinct patterns in the enumeration of working women. Spitalfields shows a steep drop in the percentage of women who were married or widowed and in employment, and a fluctuating level of employment in single women. In Camberwell, however, the variations are very slight and the working patterns show little change over sixty years beyond a slow but steady increase in the percentage of women who were married or widowed and had an occupation. Bethnal Green, located a little further from the centre of the city than Spitalfields, shows a correspondingly less marked decline in married and widowed women's work and less fluctuation in the proportion of single women working. Finally, Saffron Hill shows an intriguing pattern with both married and widowed women who worked, and also those who were not employed, dropping as a proportion of the women in the district, whilst single women formed a greater percentage of the women working recorded in the district overall. This might suggest a decline in marriage in the area. However, the figures for 1901 may skew the result due to the dramatic decline in population in the area between 1891 and 1901. At its peak in 1861 the enumeration district was home to over 700 women.
By 1901 however only 102 remained, and nearly half were unmarried -a significant, but unexplained, change in the demographic structure.
As regards the types of job shown in the census, if the CEBs were not recording casual and irregular women's work as has been suggested, then it might be expected that a relatively small number of occupations would appear: perhaps some tailoresses, a few shop-keepers, laundresses, charwomen, the odd nurse or midwife, a teacher or two per school, and so on. What is actually seen is totally different. Over the sixty years (1851-1901) 178 different job titles are given in Bethnal Green, and a similar number in  Spitalfields. These occupations are as diverse as waterproof maker, sweetmeat maker, farrier and valentine maker. The usual women's jobs are all in evidence, dressmaker, charwoman, domestic servant, and so on, but in 1881 alone thirty-six job titles appear for the first time, including velvet-coat maker and paste fitter. 58 In addition to these, a high proportion of the occupations recorded over the whole period examined here are what would be regarded as casual or seasonal jobs such as straw-hat maker and waterproof maker, and were based in workshops scattered around the area. In Saffron Hill a truly staggering 218 different occupation titles are given, including 'milk business', looking-glass maker, gold-chains maker, gold-leaf maker, and japaner. (These last probably relate to the district's proximity to Hatton Garden, the jewellery district). Camberwell stands out for sheer novelty, being home to 181 different occupation titles, including burlesque actresses, some dancers and an electrical primer tester. If such specialized occupations are being recorded in the CEBs, then this would suggest a detailed recording of women's work. It would appear, therefore, that the census, far from failing to show specialized labour patterns, shows them very clearly. It is not simply the changing patterns of employment that appear in the census -the coming and going of different occupations over the years -but the fact that occupations known to be seasonal in nature, and not necessarily taking place at the time of year when the census was taken, appear in the CEBs. For example mantle making is stated by Clementina Black to be a seasonal occupation with serious 'slack' times (late winter and spring), when minimal work was available. 59 However, in nearly every CEB studied, numerous mantle makers can be seen. The same can be said of other casual and seasonal occupations such as wood cutters and straw bonnet makers, charwomen, rag sorters and hawkers.
Removing single women from the analysis and showing only married and widowed women, provides an even more dramatic picture, as seen in Graph 6. It is still Camberwell here that has the smallest proportion of married and widowed women in work, one third that of Spitalfields in 1851. However, the number slowly begins to rise as the number of working-class women in the district increased. Saffron Hill shows its mixed fortunes as the number of married women carrying out paid work fluctuates, but forty percent of them are still recorded as working in 1901. For Bethnal Green and neighbouring Spitalfields, however, the graph shows a steep fall in married women stating an occupation -a decline that proved catastrophic for the families concerned.
Bethnal Green and Spitalfields illustrate how the census is sensitive to changes in a district, and how well women's work must have been recorded. These two districts were home to silk weavers, originally based in Spitalfields, then spilling over into the western end of Bethnal Green. Most of Spitalfields was home to weavers, which explains why the proportion of women in work was so much higher than in neighbouring Bethnal Green. However, between 1830 and 1870 silk weaving in the area collapsed. 60 By 1860 weaving was in decline but families were still hanging on in hope of a reprieve. The Cobden free trade treaty with France of that year sealed their fate. An examination of the CEBs for both districts shows that in the early years, 1851-61, they were heavily populated with silk weavers and winders, and entire families were involved in the trade. By 1901, however, not a single female weaver was listed in the Spitalfields enumeration district concerned, and only a very small number in Bethnal Green. 61 The picture is of an entire workforce wiped out by changing trade patterns, its fate mirrored perfectly in the CEBs, and clearly evident in the steep decline in occupied women in the graphs.
In most instances, the results found in the CEBs mirror known changes in circumstance in a district, and also correlate with the figures given in other sources, notably the Women's Industrial Council reports 62 and Charles Booth's Life and Labour of the People in London. In Booth's samples and notebooks for streets which were coloured black, dark blue, light blue and purple in his maps of social class distribution (and therefore predominantly working-class), we find that, in the poorer streets, he records on average around thirty percent of married and widowed women as being employed. 63 When comparing this figure with the graphs presented here it is immediately noticeable that for 1891 (and 1901) the proportion of married and widowed women in employment in the areas studied also averages around thirty percent. Therefore, the numbers given in the census for married and widowed women in work in the districts studied were consistent with those found by Booth. Whether all women who were in work on the night of the census stated their occupation can never be known. The London censuses do, however, match closely other sources from the time, and they also clearly show the myriad positions, both regular and casual, that were open to women. Whilst they may not offer an exhaustive record of every woman in an occupation, they do offer an extremely detailed picture of the socio-economic status of the districts at the time, and are sensitive to the changes experienced by those living and working there.
To establish the accuracy of the enumeration of working women in London, however, it is appropriate not only to consider whether the patterns of occupational enumeration shown in the graphs mirror the changes in occupational opportunities known to have been available within the city, but also briefly to compare the results of this analysis with the results of the corresponding study discussed here, of the enumeration districts in the provincial towns. In Norwich, on average, fifty-one percent of women recorded as having an occupation were married or widowed, while for Ipswich and Colchester the figures are thirty-eight and forty percent respectively. In London the average percentage of employed women who were married or widowed was fifty-five percent in Spitalfields, forty-seven percent in Bethnal Green and Saffron Hill, and thirty-four percent in Camberwell. The percentage of married and widowed women recorded as being in employment in the London districts is thus very similar to that recorded for the provincial towns. Allowance has to be made for the possibility that women's work was under-enumerated in the CEBs for both sets of communities, but the correlation between the types of work recorded in the census and the nature of the work known to have been available suggests that it is far more likely that both are actually recording women's occupations rather well. Further research is necessary to be able to say with certainty that this was the case, but this study has shown so far that in enumeration districts spread throughout London and the Eastern counties, the census enumeration of women's work mirrors closely what we know of the history of the districts concerned. * * * The conclusion to be drawn from the above analysis is that although there are plainly problems with the nineteenth-century censuses in regard to the analysis of women's employment, these do not necessarily invalidate their use. As Higgs pointed out in 1987, and in subsequent works, the changes to the classification of women's occupations in the Census Reports reflected ideological shifts in the understanding of women's role in society. 64 However, on the evidence presented here these ideological limitations do not necessarily affect the raw returns in the CEBs. Perhaps too much has been made of some isolated studies showing, or at least purporting to show, under-enumeration. The CEBs certainly do not pick up all casual, seasonal, and other irregular employment. However, such work was not confined to women. The analysis undertaken here requires more follow up, but with new sources becoming available it will soon be possible to examine women's work in the census at the national level. The resultant findings will be presented fully in an intended monograph. In the meantime there appears no good reason why historians should not make full use of this extraordinary source for reconstructing the social and economic roles of women in late nineteenth and early twentieth-century England and Wales.

Amanda Wilkinson is currently the Eastern ARC Research Fellow in Digital
Humanities at the University of Essex, where she is examining women's work in the census using the Integrated Census Microdata (I-CeM) dataset.

NOTES AND REFERENCES
Readers may be interested to read 'Labouring Women; a reply to Eric Hobsbawm', by Sally Alexander, Anna Davin and Eve Hostettler (HWJ 8, autumn 1979, pp. 174-82), an early comment on the nineteenth-century census as a record of married women's paid employments. Census enumerators, we argued, overlooked their employment because it was often casual, seasonal, part-time and intermittent and patterns of employment varied with the life-cycle. Much home-work was done (as it still is) in the home, in the interstices of domestic labour, and might be invisible in official sources, as it was sometimes to husbands and children. These claims were based on the authors' own research (some published, some in process) on agricultural work and on Victorian London, and were a riposte to the claims of much labour history that married women in the nineteenth century did not work. Our sources -besides those cited by Edward Higgs and Amanda Wilkinson (Henry Mayhew, Parliamentary Papers, Select Committees, Charles Booth, Clementina Black for instance) -included memoirs and oral histories and our own research was buttressed by the work of members of the London Feminist History network, by the Political Economy of Women group, and by a cluster of feminist historians gathered around Leonore Davidoff and Paul Thompson in Essex. The extent of women's employment, its reach beyond the categories of the Census, its value for family survival and economic productivity and for the shape and timing of the Industrial Revolution (as Maxine Berg and Pat Hudson so thoroughly documented) were our purpose and project; we were arguing for the inclusion of women's work in Labour and Social History and for the economic and political value of women's labour in and outside the home, especially that of married, widowed and older women.
Sally Alexander and Anna Davin 1 For a description of processes at each census and the Census Reports created, see Edward Higgs, Christine Jones, Kevin Schu¨rer and Amanda Wilkinson, The Integrated Census Microdata (I-CeM) Guide (http://www.essex.ac.uk/history/research/ICeM/documents/icem_ guide.pdf), pp. 1-152. In 1801-31 the forms (only headcounts) were filled in for parishes by local overseers. For 1841-1901, householders (they didn't have to be male but often were) filled out household schedules, which were then copied, and probably simplified, by enumerators into