Female entrepreneurship: business, marriage and motherhood in England and Wales, 1851–1911

This article o ﬀ ers a new perspective on what it meant to be a business proprietor in Victorian Britain. Based on individual census records,it providesanoverviewofthe fullpopulationof female business proprietors in England and Wales between 1851 and 1911. These census data show that around 30% of the total business population was female, a considerably higher estimate than the current literature suggests. Female entrepreneurship wasnot a uniformexperience.Certain demo-graphics clustered in speci ﬁ c trades and within those sectors employers and own-account proprietors had strikingly di ﬀ erent age, marital status and household pro ﬁ les. A woman ’ s life cycle event such as marriage, motherhood and widowhood played an important role in her decision whether to work, the workavailabletoherandtheentrepreneurialchoicesshecouldmake.Whilemarriageandmotherhoodremovedwomenfrom thelabourforce,theyhadlessofane ﬀ ect on their levels of entrepreneurship. Women who had young children were more entrepreneurial than those who had none, and entrepreneurship rates rose with the arrival of one child and continued to rise the more children were added to the family.

In the past decade a thriving literature on female entrepreneurship in Victorian Britain has emerged. 1 The second half of the nineteenth century has long been seen as the period during which the division of private and public spheres solidified, with women withdrawing from business to become homemakers, constrained by laws of coverture and social conventions. 2 However, it is increasingly recognized that the division of spheres misrepresented the nature of female economic activity; in reality many women had to, and did, participate in the labour market, as business proprietors as well as waged workers. 3 Part of this misconception derived from a gendered understanding of the concepts of work and entrepreneurship. 4 As a result of masculinized understandings of entrepreneurship, male entrepreneurs seem more visible in the historiographies of the industrial revolution, with women who ran businesses portrayed as eking out a living from micro-enterprises. 5 This ignores two important points: first, despite the focus on large industrialists or merchants, the vast majority of male businesses were also small. Second, entrepreneurship is not constrained to large, successful or innovative businesses. Running a small business as a sole proprietor required entrepreneurial skills such as anticipating demand and supply, client acquisition and management, book-keeping and, most importantly, responsibility for the enterprise and bearing its risks. Indeed many leading modern commentators include all self-employed as entrepreneurs, since they are the recipients of incomes from business proprietorship as defined in national accounts in both modern and historical studies. 6 David Blanchflower refers to this as the simplest definition of entrepreneurship. 7 It is also the most inclusive, ranging from the great business leaders to the humblest 'necessity' or 'survival' entrepreneurs. This draws from the earliest historical theorization by Richard Cantillon, who argued that entrepreneurs are defined by those accepting the risk of their trades, whilst waged labour passes risk to others: 'all the inhabitants of a state . . . can be divided into two classes, undertakers [entrepreneurs] and hired people.' 8 This article follows this inclusive approach to entrepreneurship, defining entrepreneurs as all employers and sole proprietors working on their 'own account', essentially including all self-employed. This is possible by using, for the first time, the full individual census records to provide an overview of the population of female business proprietors in England and Wales between 1851 and 1911. This article examines their numbers, the sectors in which they were most prominent, and the impact of age, marriage and motherhood on entrepreneurial activity. As we will show, the level of female entrepreneurship considerably exceeded current estimates in the literature, showing that women comprised a far higher proportion of the 3 X. You, 'Women's labour force participation in nineteenth-century England and Wales: evidence from the 1881 census enumerators' books', Economic History Review, early view (2019), DOI: 10.1111/ehr.12876. business population than previously argued. Considering the limitations of the census enumeration of female occupations, the true number of female business proprietors was likely to have been even higher, which offers a new perspective on what it meant to be a business proprietor in Victorian Britain.
The article is structured as follows. The initial two sections discuss the data, addressing firstly, the issues regarding census enumeration of female occupational data, and secondly, the method of identifying proprietors and the database which underlies this study. Next, we present the numbers, sectors and trends for female entrepreneurship between 1851 and 1911, before examining the most common businesses for women in more detail. Finally, we use a model to examine the effects of marriage and of having small children on women's participation in business. This article shows that women's entrepreneurship was profoundly shaped by their economic and demographic context. The choice to start a business was available to women in this period, but it was a choice many were forced to make because of difficulties of accessing waged labour, or out of necessity, driven by a need to survive, by seeking added income to support a family. This was less often about risk-taking, innovation or other factors that have characterized previous discussions of the heroic nature of entrepreneurship. The small-scale, uncertain, often necessity-driven entrepreneurship of women shared much in common with the majority of male businesses in this period and was more characteristic of Victorian and Edwardian business proprietorship than the male-dominated, large-scale entrepreneurship discussed in much case-study focused business history.

Women's work and the census
Locating female entrepreneurship is a complex matter with findings dependent on the primary sources selected for use, and there are significant variations in how the contents of each source reflected the actual activities of women. 9 Many women's business activities have been obscured because they took place within the same space as their domestic activities and were seen as an extension of these. 10 Some businesses were occasional and irregular: married women took in laundry if their spouses were unemployed, becoming self-employed subsistence entrepreneurs during slack periods in the male seasonal labour cycle. 11 Many women advertised their business through trade cards or were listed in trade directories, but, equally, cultural values regarding women and the public sphere led to others relying on word-of-mouth as a way of attracting custom, making these businesswomen less visible in the historical record. 12 This article draws on the newly available individual-level electronic census data (I-CeM) for England and Wales in order to examine the population of female entrepreneurs in more detail. 13 We have extracted and enriched all identifiable business proprietors from I-CeM, which have been deposited in a new database, the British Business Census of Entrepreneurs (BBCE). 14 Census data were captured at the household level and include individual demographic details of the complete population of England and Wales; its electronic version allows fuller evaluations than previously possible. 15 Here we use age, marital status, number and age of children, as well as the role of any spouse to examine the role of the life cycle and family in women's business activitiesaspects that, due to source limitations, have often been dealt with in a limited manner in recent case studies.
The issue of the recording of women's work in the census has been the subject of much debate. Recent and ongoing re-evaluations of the enumeration of women suggest that it was more accurate than often assumed. However, several problems remain. 16 The census collected household information via individuals' relationship to their 'head of household'. Heads were predominantly male, and a married woman was rarely head if her husband was present. In addition, the census enumerators who copied the schedules into the Census Enumerator Books (CEB), the General Record Office (GRO) clerks who processed and tabulated the information, and the officials who issued the instructions were almost exclusively male. 17 Against a background of gender relations in the Victorian age, it has been argued that preconceptions of a women's place in society biased the instructions, the responses and the enumeration of women in the census. 18 As a result, women's work is often considered to be under-represented, with married women's work particularly underenumerated. More recently, this view has been challenged by the findings 12  that in areas where many women were enumerated as working, such as in the Lancashire textile factories, this included high numbers of married women. 19 Also, as Edward Higgs has pointed out, many of the usually cited problems with women's enumeration relate to published tables which edited the census responses; these problems disappear when working with the original manuscript CEBs. 20 Scholars who have performed nominal linking with other records of female economic activity, using trade directories or wage records of employees in a mill, show that the original CEBs include the vast majority of the women. 21 While this offers confidence in using I-CeM, and thus BBCE data, which are derived from original CEBs, and for 1911 from householder's original responses, some caveats still apply. Several studies show that the CEBs are an accurate source for women's full-time and regular employment, but that the census did not record seasonal, irregular or part-time work. 22 While this holds for men as well, it is likely that women, and married women in particular, would have been more heavily affected. 23 Another issue particularly pertinent to the study of female entrepreneurship concerns women's contribution to running small home-based businesses, where the production of goods and services was naturally part of the household setting. Here, wives and other female family acted as co-workers or de facto partners but were less frequently recorded as such in surviving records.
In addition, there were several inconsistencies between the censuses over the period of study. Firstly, the questions asked by the census were not worded consistently. For instance, the 1861 instructions read 'WOMEN AND CHILDREN to be entered according to the above Instructions'namely the instructions for menand that 'The occupations of those who are regularly employed from home, or who follow any business at home, to be distinctly recorded.' 24 In 1891 this specific instruction was much more limited, just stating that 'the occupation of women and children, if any, are to be stated as well as those of men'. 25 In 1901 the gender-specific instruction had disappeared completely and, when it reappeared in 1911, the instruction was directed at 'women engaged in any business or profession, including women regularly engaged in assisting relatives in trade or business'. 26 Secondly, there was a change in the proportion of women for whom the occupation field was left blank. For single women, this hovered between 20 and 25% of the populationa trend that remained remarkably constant over the census years. Married women, on the other hand, showed inconsistencies between the early censuses and the later censuses. In the 1851-1881 censuses, the occupation field of 70% of married women was left blank, as opposed to 85 to 90% in the 1891-1911 censuses. Widows, finally, showed a more gradual increase of blank records: where in 1861 around 30% of widows over the age of 60 had a blank occupation field, this rose to 40% by 1891 and almost 50% by 1911. The increase in nonrecorded occupations between 1881 and 1891 is also evident, albeit to a lesser degree, in older men. There are no obvious reasons why there was such a sudden increase in blank occupations for certain groups, but it is likely that the new census question on employment status introduced in 1891whether someone was an employer, own-account proprietor or workersharpened perceptions of what an 'occupation' meant. People who had a less defined occupation or worked part-time may have elected not to fill out the occupation question, something which would be more common in ever-married women and older men, and less common for single women and younger men.
A final issue relates to changes in the enumeration of married women. Here the wife's work can be hidden under the occupational descriptor of her husband, for instance 'carpenter's wife', which was used as much as a social status descriptor as an occupational title. While a proportion of these were partners in the household business, others were not economically active. These proportions varied by enumerator; some enumerators reported every married woman as 'wife', and some only gave the description to a fraction of married women. 27 In addition, there was considerable variation of women thus enumerated between the censuses, with 1891 and 1901 having far fewer 'wives' than the other years. In the absence of a way to reliably distinguish between genuinely economically active women and those who were enumerated as a social status, and without a possibility to adjust for them in 1891 and 1901, they have been removed from the economically active population in the following analysis. Whilst we acknowledge that this removes a number of married women who would have been active partners in their family firm, the variation of reporting between and within censuses makes this necessary to avoid distortions. It should therefore be recognized that the number of married businesswomen in the following analysis represent a minimum.
Despite these issues with the census, as Higgs and Amanda Wilkinson argue, this should not invalidate its use. 28 Those who criticize the use of the census for the study of female occupations often recommend other sources; for entrepreneurship this is usually trade directories. However, directories suffer from similar omissions, particularly with regard to married women and, as this article will show, identify lower numbers of female entrepreneurs than the census data. In addition, there is reason to believe the enumeration of entrepreneurs in census data was better than the enumeration of workers. 29 For instance, a lodging-house proprietor, even if only requiring a few hours activity a day, was more likely to be enumerated than a woman spending a similar number of hours working in a family member's business if and when required. The census data reveals a larger sample of female business proprietors than found in other sources, and allows us to define entrepreneurship for women on comparable terms to men: as proprietors of their own business of any size. Despite the known issues particularly regarding the occupations of married women, as a source of information about business proprietorship the electronic census data are unrivalled.

Locating businesses in the census
As the census was aimed at counting individuals rather than businesses, the I-CeM data required extensive manipulation in order to extract all entrepreneurs. The BBCE includes as entrepreneurs all those who were responsible for running their own business, regardless of its size. 30 We use here the key definition of entrepreneurs as those who received their income as a residual of their profits after deducting costs and thus took on a risk, as opposed to workers who received a wage. This definition incudes all self-employed people, both business proprietors who employed other people (employers) as well as those who were run by one person: that is, self-employed sole proprietors, or, in the phrase used in the nineteenth century, 'own-account' businesses. The latter type of business was smaller and could also be more precarious and based on necessity because of lack of alternatives in waged work, but it represented an undertaking nonetheless. It is also the format of the census enquiry from 1891, which is the format that has continued in modern censuses. Hence, the analysis developed here allows direct comparison with modern definitions.
The census data on entrepreneurs in the BBCE relies on individuals' self-reported employment status. This was reported in different ways. The 1851 to 1881 censuses asked employers to report their workforce size. After 1881 this question was discontinued and the 1891 to 1911 censuses asked respondents, in addition to their occupation, to indicate whether they did this as workers, employed others or worked on their own account, allowing us to identify the self-reporting business-owning population. However, it was known by contemporary census administrators that there were nonrespondents, although no attempt was made to clean and correct the data. 31 Modern published census tables are weighted and adjusted for significant non-response and other biases, which often occur for the sort of occupational questions that are the basis for identifying entrepreneurs. 32 I-CeM now allows similar adjustments to be made to the historical data. For the majority of non-respondents, their status can be confidently determined based on their occupational descriptors, such as domestic servants as waged status. For the remaining non-respondents, the appropriate data weights to compensate for non-response were calculated using a logit regression to estimate the probability of non-response based on its strongest correlates: gender, relationship to head of household, and occupation. 33 A second and more major adjustment is needed to align the 1851-1881 responses to those from 1891. Table 1 shows the number of employers, broken down by gender, who responded by stating their workforce in 1851, 1861 and 1881, and those who self-reported as employers in 1891, 1901 and 1911 (which includes weighting for non-responses). 34 It is obvious from the discontinuity between 1881 and 1891 that there were different levels of response to the earlier census instructions, and that this was affected by gender. There were several aspects to this. Firstly, the 1851-1881 census respondents reported their workforces in line with two instructions: one was specifically targeted at farmers, who were asked to report their acreage and their number of labourers. The second instruction asked 'masters' in trades, manufactures or other business to report their workforce. Not all employers considered themselves masters, and there were separate instructions for certain professions, which made it ambiguous whether they were supposed to respond to this question. The imbalance between reported employees and the number of self-reported workers indicates that not all employers followed the instruction. In addition, the language used in the 'master' instruction was gendered, and far fewer women responded to this compared to the instruction for farmers, which had much higher female response levels. Finally, the 1851-1881 censuses only explicitly identified employers, while women who ran businesses mostly did so on own account. The breakdown of employment status in the later censuses shows that the majority of businesses, both male-and female-run, were own account, but women were most active in sectors where this was particularly the case, such as maker-dealing, personal services and small-scale retail. For these reasons, the BBCE supplements the extracted entrepreneurs from the early censuses to align them with the later census questions. This is the basis for the data used in this article. For everyone apart from farmers, a logit regression based on the 1891 or 1901 census data (weighted for non-response) was used to distinguish between workers and entrepreneurs (either own account or employers). This regression used the variables: age, gender, marital status, relationship to head of household, occupation, population density and the number of servants in the household. This generated coefficients for each economically active individual in 1891. These coefficients were then applied to the 1881 census data to generate scores that gave the probability that any given individual was an entrepreneur. This was repeated using the 1901 weighted data. Furthermore, two ratio back-projections were calculated. This process gave four supplemented totals for each occupation category, and the most appropriate total was chosen based on trends and secondary literature. These aggregates were then populated with individuals; first, any people identified directly as employers or masters from their census responses were included, the difference between these and any deficit from the reconstructed total was then supplemented with individuals based on their probability scores from the logit regressions. This process was then repeated to split the entrepreneurs into employers and own account, and then repeated for 1851 and 1861. The supplementation process allows the identification of individuals who were either definitely entrepreneurs (those who responded directly as employers or masters: the extracted) or who were most likely entrepreneurs based on their characteristics using the logit estimates (the supplemented). 35 It produces an estimated population of entrepreneurs for 1851-1881 that is aligned with the entrepreneurs identified directly from the census by the 1891-1911 questions. 36 For farmers a more direct method was used. All farmers responding as self-identified employers again provide the starting point to enable categorization as entrepreneurs. The remaining farm entrepreneurs were identified from those who provided only acreage data. These were split into employers who had large acreage who should have responded as employers, and own-account proprietors who had small acreage. A logit regression was used to identify an acreage cut-off for each Census Registration District above which a farmer would have had to employ labourers. 37

Female entrepreneurship
The process to align the census responses across the censuses allows the female and male entrepreneurs to be fully identified and compared with each other. Table 2 shows their numbers for the whole period 1851-1911; for 1851, 1861 and 1881 based on supplemented employer and master responses, and for 1891, 1901 and 1911 based on the actual employer and own account numbers weighted for non-responses.
The proportion of female entrepreneurs remained fairly constant at just under or around 30% of the total business-owning population. The drop of  35 This approach relies on the assumption that, for the supplemented, the demographic and household characteristics of entrepreneurs in a particular occupation in a particular location did not change substantially over the 40 years from 1851 to 1891; but the directly extracted are not affected by this. 36 A full discussion of the extraction and supplementation methods deployed to create the database of business proprietors in BBCE is available as a working paper: see R.J. Bennett own account in 1911 mirrors an overall drop of the self-employed business-owning population, mainly due to increasing mechanization of maker-dealer activity and business concentration in larger enterprises; however, the number of female employers increased. 38 These estimates can to some extent be benchmarked against previous case studies. However, they demonstrate that, despite any uncertainties from the supplementation process and possible undercounts of married women, these numbers are substantially higher for all years than most previous estimates of female entrepreneurship in Britain, which estimate female businesses at between 5 and 10% of the total. 39 The comparison with previous studies can be extended in more detail. Much of the recent scholarship on female entrepreneurship in Britain has been based on urban case studies, mainly using trade directories and insurance records. We have tested selected locations, which show that compared to the census, directories systematically under-record married women, women in certain sectors and multiple entrepreneurs in a single household. In addition, many trade directories only stated an initial rather than full name, which prevents identification of gender. Burnette's work on mid-nineteenth-century trade directories in Birmingham shows that while 11.8% were identifiably female, another 8.9% were gender unknown, meaning that the potential population of listed women could be over 20%. 40 These differences explain why Jennifer Aston's estimates of female entrepreneurship in Birmingham and Leeds based on directories, ranging between 3.3 and 8.2%, are so much lower than the census, where respectively 26 and 35% of entrepreneurs were female; although the trends in female entrepreneurship she gives match those based on census data. 41 Fire insurance records, as used by Alison Kay in her study of female entrepreneurs in nineteenth-century London, have their own inherent bias in that businesses with higher capital needs were more likely to be insured, and businesses taking place at home with few additional business assets are less visible in policies. The type of business often run by men falls in this first category, while women ran enterprises of the second kind, which explains Kay's estimate that women were not likely to account for more than 10% of total businesses. 42 In addition, some trades would have been more vulnerable to fire, and be more likely to appear in the records. For instance, chandlers appear in Kay's top 10 businesses for both men and women, but do not even make the top 100 of entrepreneurial occupations in the census. 43 The differences between the sources can be compared for 1851. Figure 1  displays  created by Kay, and adapted by Aston. 44 The census occupations have been grouped in the same manner for this chart in order to compare results; in the rest of this article, a different sector classification is used. In order to make results comparable, the census results have been restricted to urban locations only. Two census estimates are given: firstly, census employer responses for non-farmers who responded to the 'masters' instruction; and secondly, the extracted and supplemented data estimates which include all female entrepreneurs whether employers or own-account proprietors. The supplemented estimates provided in the BBCE show significant advances over alternatives. The employers responding to the census instruction show similar proportions of female entrepreneurs in food and textiles as the fire insurance records and the Birmingham trade directory. There were far fewer non-textile retailers identifying themselves as employers in the census, and considerably more miscellaneous, which is not surprising as these include all non-textile manufacturers and the respondents answered a question aimed at manufacturers. Non-textile retail was low in the census respondents because most of these businesses were run by own-account proprietors. There were almost no school employers; where these existed they would have been own-account proprietors. The extracted and supplemented data, on the other hand, has a considerably larger proportion of textile entrepreneurs than any other source. As evident from the nonsupplemented data employment-type breakdowns in the 1891-1911 censuses, many of the trades included in textiles were run on an own-account basis, and it is clear that these were not picked up in early census responses, and were also largely absent from the trade directories and fire insurance records. London's fire insurance records show that 15% of female entrepreneurs were dressmakers or milliners, although in the censuses they consistently accounted for well over 30% of female entrepreneurs. 45 Laundresses did not even appear in the 10 most common occupations in the insurance records, but consistently accounted for 10 to 15% of female entrepreneurs in the census. Finally, evidence relating to farmers is also strongly affected by the source used. Trade directories were generally urban in nature and farmers were rarely listed, thus omitting a large category of entrepreneurs. The farming sector's size meant that female farm entrepreneurs, while comprising only a small proportion of all farm entrepreneurs, still amounted to a considerable part of the female business population. The urban focus inherent in the use of trade directories obscures this important sector.
International studies based on census records can also be used to benchmark our estimates. These closely match our estimates. In Canada, the 1901 census showed 30% of business proprietors were women. 46 In Belgium, census data between 1880 and 1910 showed that 34% of businesses were female, while German official statistics showed around 25% female businesses between 1882 and 1907. 47 In addition, if we look at the population of shareholders in England and Wales, who effectively owned part of an incorporated business, we find similar proportions of female involvement as in the census. Female shareholding in a range of businesses rose from 24 to 34% between the 1880s and the 1910s, while similar numbers were found for shareholders in various banks. 48 While the overall female entrepreneurship share hovered around the 30% mark, there were vast differences between sectors. Table 3 shows the percentage of female entrepreneurs as part of the business population by 13 sectors over six census years between 1851 and 1911. In several sectors, female entrepreneurial participation was consistently low; construction, transport, professional and business services, mining, finance and commerce, farming and agricultural produce processing and dealing were male-dominated sectors in this regard. In personal services and maker-dealing, on the other hand, women constituted the majority of entrepreneurs for most census years. They also formed a significant minority in refreshment, retail, food sales and manufacturing.
Within these broad categories, there was further variation. The high proportion of women in personal services was mainly caused by the large numbers of laundresses, a trade that was over 90% female. Other personal services occupations were much more male-dominated; hairdressing, for instance, was over 95% male. Even the most 'masculine' industries such as construction were not uniformly male and included some small occupations in which female entrepreneurs were common, such as whitewashing, where almost a quarter of entrepreneurs was female. Retail is often considered a sector that employed many women; yet it was a diverse sector that offered female entrepreneurs opportunities in some, but not all sub-sectors. Most women were, unsurprisingly, concentrated in clothing-related retail, but general shopkeeping, hawking, stationers and artificial flower shops were often headed by women as well. Textile raw materials, on the other hand, were often sold by men, as were skins, leather and metal. Food sales saw a similar divide between male and female occupations, with butchers being over 90% male, while confectioners were predominantly female. An example of an extreme gender split can be found within the refreshment sector, where women made up almost 90% of all lodginghouse keepers, but less than 20% of inn and hotel keepers.
non-retail manufacturing trades outside textiles and food. 49 In the census, however, non-retail female entrepreneurs not in textile or food-related sectors comprised no more than 3% of their total. This difference is, once again, explained through the difference in data sources: some of these masculine trades had higher capital needs, and were more likely to obtain an insurance policy, whereas some of the main female occupations, dressmaking and laundry, required little capital. Table 4 shows the top 10 occupations by census year, based on the most detailed level of occupational coding available in I-CeM. These occupations together consistently accounted for around 70% of female entrepreneurs. Some trades remained important throughout the period: dressmakers consistently came out on top with the related clothing manufacturing occupations of milliner and shirt-maker/seamstress listed in the top 10 in most years. Laundresses were also prominent, while grocers, innkeepers and shopkeepers also consistently made the top 10. Other sectors were subject to some change: the straw plait manufacture industryin the top 10 in 1851 and 1861collapsed towards the end of the century and accounted for less than 0.1% of female entrepreneurs by 1901. The decreasing importance of agriculture to the British economy during this time can be seen in the decline of farming, dropping from third most common occupation to fifth. Education, as schoolmistresses or music teachers, appeared in the top 10 in the second half of the census period, highlighting the increased feasibility of setting up schooling or private tuition as an entrepreneurial choice for women. Lodging-house keeping increased in importance over time, and while some of this was driven by changing definition of what constituted a lodging house, the trend is clear. 50

Key female businesses
Since female work and entrepreneurship were so skewed towards a few occupations, the demographics of these key occupations drove much of the trends. It is therefore necessary to pick out the specific driving forces for different groups of women, which reveal their varied experience of entrepreneurial life in Victorian Britain. Figure 2 contrasts employment type over the course of the life cycle for the main occupations of female businesses. The data is based on 1901, a year chosen because it has explicit recording of employer status rather than using supplemented data and is the most accurate employment status information before the decline in self-employed businesses in 1911; however, other years showed very similar breakdowns. Dressmaking displays a demography dominated by young workers, but also sustaining many women throughout their lives. Contrast this with lodging-house 49   keepers where very few women were either young or workers. This industry was dominated by women at mid-life, the vast majority working on their own account with substantial numbers of women still running lodging houses at advanced age. Both laundresses and female grocers show a peak at the youngest group, with a large proportion of workers, followed by a decline in the 25-34 age group before a gradual rise in numbers again, driven by increasing entrepreneurship. The young workers, who were likely to be employed in a family business, dropped out of the economically active upon marriage, before re-entering as businesswomen in their own right. Finally, farmers display a demography also seen by female entrepreneurs in the 'traditionally masculine' industries: very small numbers of workers and young people, and a large proportion of employers at higher age. This was mainly driven by widows, who continued a business after their spouse died.
Of the other top 10 occupations, those that were in the same sectors often followed similar trends. Not surprisingly, milliners were similar to dressmakers, although their entrepreneurship rates were much lower: overall 77% of milliners were workers, versus only 49% of dressmakers. This difference was particularly stark in the youngest age groups, with only 10% of young milliners being entrepreneurs, while over 30% of dressmakers ran their own business at the same age. Seamstresses and shirt-makers were even less entrepreneurial at a young age; while their overall percentage of workers was the same as for milliners, only 3% of seamstresses in the youngest age band were not workers. Dressmaking therefore stands out as a trade that allowed women to set up their own business from a young age, even compared to other parts of the clothes-making sector. As Charles Booth pointed out, little capital was required to set up a dressmaking business. 51 However, the trade varied from London society showrooms to smaller establishments where clients brought in fabrics or department store dresses for alteration, and had different associated start-up costs. 52 As a sub-sector, it increased in importance between 1851 and 1901 in offering women business opportunities. In 1901, at its peak, well over 70% of dressmakers aged over 35 ran their own business, while for those under 35, only single women under 25 were more likely to be workers rather than entrepreneurs. The dressmaking industry declined in 1911. The reasons for this are complex, and can be found in a combination of factors: the availability of sewing machines at home, leading to a rise in amateur dressmaking for the family; new organization systems that transformed dressmaking from a craft into a more industrialized system; and the rise of the department store. 53 It should be noted that milliners, whose craft could not be easily mechanized, did not experience a similar drop in 1911.
Of the main female occupations, dressmaking was most dominated by never-married women, and a career in dressmaking allowed a single woman some independence: 13% of single entrepreneur dressmakers were heads of households, while only 3% of worker dressmakers headed their own household. This opportunity of an independent life was deemed one of the more compelling reasons to choose a dressmaking apprenticeship over other occupation options, as indicated by the introduction of an 1843 Guide to Trade for dressmaking, which also mentioned that a dressmaker would be able to have a house of her own. 54 Both employer and own-account dressmakers under the age of 35 were slightly more likely to be married than worker dressmakers, a difference that almost disappeared in the older age bands. Clearly, for young dressmakers, marriage either enabled them to set up their own business or made it more difficult to work for someone else.
Laundry work was considered a lower-class occupation and often seen as one of the sweated trades. 55 Booth noted the large proportion of laundresses that were married women, and indeed, in contrast to dressmakers, laundresses were the group of female entrepreneurs most likely to have been married. 56 Laundry workers were reported by their employers to be often the main wage earner, supporting their families when their husbands were (temporarily) out of work. 57 Marriage was even more common for business-owning laundresses than for workers; however, entrepreneurial laundresses had a similar domestic set-up as workers and were predominantly married to general labourers, agricultural labourers, or those active in the building trades. Running a laundry business was evidently a strong opportunity for a married woman from an early age. It required little additional skill and equipment beyond what was already needed for the household, and the work could be fitted into the domestic routine, was often available all year round and offered a degree of flexibility. Laundry proprietors were said to complain often about the unpunctual and irregular hours worked by their employees, who worked around their home duties. 58 Setting up a small business, therefore, would allow a woman to maintain these hours without having to negotiate with employers. Female-run laundry businesses were particularly concentrated in places containing many temporary residents, such as spa towns, sea resorts, ports and university towns. Areas with the highest proportions of female-run laundries by population included Poole, Bath and Cambridge. While London has been identified as an area with many laundresses as well, it was less entrepreneurial for this trade, with higher proportions of worker laundresses. This higher visibility of employee laundresses in the metropole may have contributed to the image of laundresses as one of the sweated trades, obscuring its important role offering women entrepreneurial opportunities. 59 The occupation most dominated by women proprietors, however, with almost no workers, was lodging-house keeping. As a profession lodginghouse keeping accounted for between 4 and 6% of female entrepreneurs over the period, mainly own account, with a rise to 8% in 1911. In contrast to laundresses, it was mainly an occupation for women who lived without a man. Lodging-house keeping had some of the highest rates of young widows (those under the age of 35), and the second highest rate of nevermarrieds after the dressmakers. In addition, it had some of the highest rates of women married with an absent spouse. This was a slightly ambiguous term. While in many cases it indicated a married woman whose husband was at a different location on census night, for instance on business or family visits, it also included single or separated women who claimed married status as a mark of respectability. Indeed, lodging-house keeping was seen as way for women to earn a living without compromising social status and was associated with domesticity and providing moral guidance. Grocers made up between 3 and 5% of all female entrepreneurs, and their characteristics reveal some of the largest differences between entrepreneurs and workers. Female grocery proprietors were much more likely to be married and, while the most numerous occupation of their husbands was also that of grocerusually indicating a de facto partnershipthere were substantial numbers of female grocers married to farmers, coal merchants and other food sales occupations. Further indication of the importance of the family as a source of labour and/or partnership is the fact that employer grocers were more likely to be widowed than ownaccount grocers, indicating that a grocery could not be run alone and after the death of one spouse the remaining one had to employ help.
Farming, while consistently appearing in the top 10 female entrepreneurial occupations, was a very male-dominated trade. Over 90% of farming entrepreneurs were male, and it was the major entrepreneurial occupation for men throughout the period. However, a lot of female involvement in farming was hidden due to census questions asking that farmers' family members be returned as 'farmer's son' or 'farmer's daughter', even if they worked on the farm. 60 Numerous wives of farmers were returned as 'farmer's wife'. It is clear, however, that many were actively involved in the farm: married male farmers with small farms were more likely to be own account rather than an employer, since they could rely on occasional labour from their household. 61 In addition, very few of the female entrepreneurial farmers were married, unlike the grocers. Female farmers displayed a pattern usually observed in male-dominated trades: mostly older, and often widowed. In addition, the majority of female farm employers had farm workers living in their household. It seems probable that they had been involved in the farm business throughout most of their marriage and continued to run their late husband's business. Many were widowed from a young age: almost 40% of female employer farmers were widowed by the age of 35.
One aspect that the main entrepreneurial occupations for women had in common is that they could be performed using the home as the workplace. The ideology of domesticity and its relation to female roles extended its reach to what was considered respectable work for a woman. Since both waged work and entrepreneurship were seen as male, women who participated in the labour market either as workers or business proprietors were already operating to some extent outside their supposed sphere. 62 In many ways this was mitigated by certain trades being considered feminine or respectable, either by operating trades that were mainly for women by women, such as dressmaking, or by connecting to the female domain of the private sphere. 63 Of the main sectors in which women ran businesses in Victorian Britain, laundry and lodginghouse-keeping relied on skills that were used in the household as well and, while dressmaking was a skilled job with an apprenticeship, most women were taught basic sewing skills and it was seen as an acceptable pursuit for a woman. Another aspect of trying to maintain a separation of spheres, at least physically, was working or running a business from the private sphere of the home. The 1901 and 1911 censuses included a question on whether an individual worked at home. Table 5 shows the proportion of men and women working at home in 1901, indicating that women were more than twice as likely to work as proprietors at home than men. The 1911 census showed a similar proportion.
The possibility of working at home was largely sector dependent, with the majority of entrepreneurs in maker-dealing, personal services, food sales and refreshment working in the same place they livedsectors which overlapped with those in which female entrepreneurial rates were highest. The lowest rates for either sex can be found in the mining and quarrying sector, again showing the relation between masculinity and working outside the household, while feminine work was performed indoors. In addition, there were stark differences between employment types. Workers of both sexes were least likely to work where they lived, whilst the majority of female own-account business proprietors worked at home. Dressmakers and laundresses, in particular, had very high rates of working at home, at close to 90%.

Marriage and motherhood
The discussion of the key sectors for female entrepreneurship suggests the role different stages of life played in business development. R.J. Morris has placed the family at the centre of his discussion of the middle-class property cycle. 64 This argued that young adults moved from earning income and loan repayment to entrepreneurial capital during the early stages of family formation, increasingly accumulating assets within the family during later adulthood, to transfer to rentier forms of income to provide for old age as well as dependent family members. For women this life cycle is particularly linked to demographic events, particularly marriage, having children and widowhood. 65 Figure 3 shows entrepreneurship rates by marital status for both women and men. Since entrepreneurship interacts between gender and marital status a logistic regression for these variables is used to test the different marginal effects of gender and marriage. This is shown in Table 6 for 1901, with other years yielding similar results. The coefficients in this table refer to the probability of an economically active person being an entrepreneur for each of these variables against the base category of married men.
Both the raw entrepreneurship rates (Figure 3) as well as the logit (Table 6) indicate that women had a higher rate of entrepreneurship than men. This is explained by the greater accessibility of wage labour for men, which made Source: see Table 2. proprietorship a less attractive option. 66 Marriage increased entrepreneurship rates for both men and women, but particularly for women. While large numbers of single women had recorded occupations in the census, they were predominantly wage labourers in sectors that offered no opportunity for entrepreneurship, such as domestic service and textile factories. Many of them dropped out of the observed economically active after marriage and the arrival of children, but those who remained were significantly more entrepreneurial than married men, who remained in the waged labour market. For men, death of a spouse reduced the probability of being an entrepreneur. For women however, widowhood increased entrepreneurship considerably.
Hence, while marriage played an important role in removing women from the labour force, it had a very different association with their levels of entrepreneurship. While previous historians have suggested that entrepreneurship was a more viable option than waged work for married women, this has not been so fully recognized previously. 67 Marriage offered a clear advantage for entrepreneurship although there are several aspects to this. In part the apparent advantage could result from better census enumeration of married female proprietors compared to married female workers, but this is likely to explain only a small part of the difference. A more important factor lies in the more limited opportunities for married women in terms of waged work, as certain professions, such as teaching, placed limitations on married women's employment. 68 In addition, women's wages were less than men's wages, making it more effective for the husband to work if only one person at a time could leave the home. 69 On the other hand, marriage brought with it added resources and it could provide access to new social networks, capital, or potential labour, all of which made starting a business easier. Finally, as marriage and a business both required certain levels of capital, a proprietor might have delayed both until a later point in their life cycle. Married people who were at home with their spouse on census night can be investigated jointly to assess the occupational behaviour of couples. Table 7 shows the employment statuses of the husbands of all economically active women in 1901. This reveals that for all groups of economically active women it was most likely that their husband was a worker. However, while over 90% of worker women were married to worker men, the marriage behaviour of entrepreneurial women was more varied, with both employer and own-account women being more likely to marry within a group whose employment type matched their own. This presence of dual-entrepreneurship marriages has previously been identified for eighteenth-century London, and while some constitute partnershipsfor instance a married couple of grocersin many other cases these were two people running separate businesses. 70 A small percentage of women had husbands who were not economically active (non-EA), implying an inversion of the male-breadwinner/female homemaker ideology. While this did not make the women sole breadwinners of their householdsince there could have been contributing children or parentsthe fact that this dynamic was more than twice as likely if the woman was an employer rather than a worker points to the possibility that her business was able to support both spouses. Additionally, in the majority of households of married female entrepreneurs, the women were more entrepreneurial than the men, with 50% of employer women, accounting for almost 3000 women, living with a worker husband. Over two-thirds of these were comprised of laundresses and dressmakers or milliners. The differences in the spouses' occupations between these key groups of female employers reveal a lot about their status in society: while the most common occupations for spouses of employing laundresses included general labourers, agricultural labourers, gardeners and construction workers, the husbands of employer dressmakers and milliners were generally clerks, company agents, drapers and shoemakers.
In addition to marriage, childbirth was a key event in the female life cycle. Childcare had a well-documented and undeniable impact on a woman's availability to engage in the labour market. 71 Of the over 300,000 women under 40 married in the two years preceding the 1911 census, 15% were economically active after marriage, but before the arrival of any children. Of those who had a child, only 10% were still recorded with an occupation. In addition, mothers' documented labour force participation rates dropped further with each child that was added to the family. The relationship between motherhood and entrepreneurship is less well researched for this period. Kay noted that just under one-third of her census-linked sample had children below the age of 14 living in their household but focused mainly on older co-resident children and business succession. 72 It is possible to take the investigation of the effect of motherhood further by linking parents with their children present in the household. This method only includes a parent's own children; it excludes other children living in a household and is not able to identify children who spent census night in a different household from their parent. Table 8 shows multinominal logit estimates of the relationship between entrepreneurship for categories of married women under 45 and number of children under five years old. The base category in this model is married female workers who had no children under five, which means that either they did not have any children (yet), or had older children who had aged out of this group. 73 The table clearly shows that women who had young children were more entrepreneurial than those who had none, a finding that remained consistent across age and occupations. In fact, entrepreneurship rose with the arrival of one child and continued to rise as more children were added to the family, although the significance level of these results is low after the fourth child as the sample size becomes very small.
The nature of the census data does not allow testing for causality, and it is likely this pattern conveyed a mix of both demand and supply factors. Parttime work or work with more flexible schedules was difficult to come by. Childcare requirements, which were predominantly left to women, could make it more difficult to comply with fixed wage working patterns, making self-employment and the flexibility to choose one's own hours more desirable.  72 Kay, op. cit., 97-9. 73 The former is more likely amongst younger women, while the latter is more likely for older women, although there was always a small proportion who married but never had children.
Present-day studies suggest that flexibility and childcare, in addition to independence and financial factors are key motivations for new mothers starting up new businesses. 74 Alternatively, having an established business could encourage adding more children, particularly if the business was undertaken at home and could facilitate childcare. As Table 5 shows, running a business offered much higher work-at-home opportunities than waged labour, and modern studies confirm the link between fertility and working from home. 75 It is likely that a combination of these factors was in play.
There are some important sector differences. At the time of the 1911 census, just under 2000 dressmakers under the age of 40 had been married for under two years. Slightly less than a quarter of them had had a child. Of those who remained without children, 9% were employers, 54% worked on own account, with the remaining 38% workers. The dressmakers who had had children were more likely to work on own account (67%), but were less likely to be employers, with only 4% employing others. As a whole, dressmakers with children appeared more entrepreneurial with only 30% being workers, but the move away from employing others in favour of being self-employed hints at smaller, and perhaps more precarious, businesses. Laundresses, of whom 1700 had been married in the two years preceding the 1911 census, followed a similar pattern, although over 90% were workers. Like dressmakers however, those who had a child were more likely to be own account rather than employ others. Alongside a wider trend of dropping out of the economically active at the birth of the first child, the small number of women who remained were remarkably entrepreneurial.
Widowhood, finally, had a very positive association with female entrepreneurship. Female entrepreneurship has been characterized as older and widowed women coming into the role of proprietor after the death of her spouse. 76 The census data confirm that the difference in average age of male and female entrepreneurs is a good proxy for the level of gendered work of a sector. For instance, in blacksmithing, an extremely maledominated sector, the age gap between male and female entrepreneurs was consistently over 10 years, implying that women only came into the trade at a higher age. This does not mean that they were less entrepreneurial than 'self-made' proprietors such as dressmakers. Many male entrepreneurs would also have benefitted from having family members in the business, and many sons followed older family members into business, whether male or female, and became partners. 77 Indeed, often these widows remained in business even after their sons were of age. The 1901 census data shows that over 30% of widows who ran a blacksmithing business had sons over the age of 21 who were also recorded as blacksmiths. Similar proportions can be found for widow painters, millers, butchers and saddlers. Rather than being placeholders, it is likely these widows had been involved as a hidden partner for at least some part of the marriage, and were the ideal person to continue the business, reflecting contributions they had made that had gone unrecorded before.

Conclusion
Many women ran their own businesses during this period, but their choice to do so, and their choices about the kind of business they ran were constrained by a series of factors. Some of these constraints can be observed in the census data: age, marital status, sector, the choice between employing others and working on one's own account. Others are invisible, in no small part due to the nature of the census as a source, most notably the impact of cultural norms about gender roles and relations. However, the census data make clear that entrepreneurship was important for many women, and that it was a choice thrust upon them by necessity and the increasingly gendered nature of the waged labour market, while also offering opportunities to those who had entrepreneurial capacity. Their entrepreneurship was thus in many ways similar to that of men: more constrained to specific sectors, but no less valid or interesting for that fact.
More than was the case for men, a woman's life cycle events such as marriage, motherhood and widowhood played an important role in her choice whether to work, the work available to her and the entrepreneurial choices she could make. While it is often argued that many female entrepreneurs were single or widowed, it is important to note that although both marriage and motherhood often had a negative impact on women's participation in the waged labour market, they were positively associated with business proprietorship. However, female entrepreneurship, as with female labour force participation more generally, was not a uniform experience. Certain demographics clustered in specific trades and, as well as women tending to be found in a relatively smaller number of sectors compared to male entrepreneurs, within those sectors employers and own-account proprietors had strikingly different age, marital status and household profiles. Our understanding of women's enumeration in the census is still developing, and there are issues that should be kept in mind while interpreting the data. However, 77 See e.g. Barker, op. cit. the data derived from digital census records is one of the better large-scale sources for female entrepreneurship that we have for the nineteenth century; and despite some undercounting of married women it captures a far larger number of female business proprietors than any other source available.
These findings should remind historians that large-scale businesses were the exception in this period, and that most entrepreneurship consisted of men and women running small businesses. The case of female entrepreneurship reveals what is lost when historians focus on large businesses and define entrepreneurship as a heroic activity based around innovation or grand risktaking; it ignores the majority of people who ran businesses in this period, both men and women, the factors that shaped their choices to start businesses no matter how precarious, and thus restricts our understanding of the structure of the economy in this period. By taking a broad definition of entrepreneurship, and by turning our attention to the women and men who were business proprietors in this period, we gain a fuller understanding of these factors. The analysis based on census data presented in this article confirms many of the conclusions reached by recent case studies in female entrepreneurship. However, it also shows that these have significantly underestimated the size of the female business proprietor population and its relation to the male one. Women in business were as numerous in the second half of the nineteenth century in England and Wales as they were in North America and Europe, and any decline in numbers began to take place during the early decades of the twentieth century, rather than during the Victorian period.