What is an ‘inactive’ person? Although the definition remains unclear, it has been the subject of a lengthy statistical classification effort. Drawing on French censuses from the late 19th century, Agnès Hirsch examines how this category was constructed and then reconstructed as the opposite of the ‘active’ population. Moral, political, economic, and technical factors are involved in the transformation of this category, which relegate the people in it to the margins of the production system.
1 Between 1891 and 1896, France’s ‘active’ population increased considerably, by around 15%. In the introduction to Volume 4 of the 1896 census, this increase was attributed to improvements in the procedures used to collect and count census data. But a look at the figures presented in the census reveals a surprising fact: it seems that nearly 2 million women entered the active population in 1896. In 1891, the census counted 4.6 million active women; in 1896, the number was 6.4 million. These figures represent 29.5% and 34.6% of the total active population respectively. Can so large an increase be explained solely by an improvement in processes? Cahen (1953) argued that while improvements in the reliability of census data constitute one explanation, other factors can also be suggested. For example, she emphasized a change in the conception of the census beginning in 1896, as well as a change in the nomenclature for different activities. Marchand and Thélot (1991) also suggested that it is important to consider the role of both process improvements and changes in conventions, particularly regarding the measurement of agrarian labour. Studies using long data series to look at how the active and inactive populations in France have changed over time have highlighted the heterogeneity of the censuses over the course of the 19th century, in terms both of quality and of the adopted definition of activity.
2 Multiple approaches have therefore been proposed to analyse changes in the activity rate over the last centuries: starting with the 1896 census, known as the first where the counting of bulletins professionnels (occupational forms: separate individual census forms used to document workers’ employment situation) was centralized in Paris; drawing on the archives of censuses carried out before the 1890s, while including various disclaimers on the nature and accuracy of the resulting information (Cahen, 1953; Toutain, 1963); constructing homogeneous data series and thereby correcting them, first by keeping in mind current definitions of activity and inactivity, and second by using pre-1896 censuses as ‘cues’ (Marchand and Thélot, 1991, p. 13). A number of questions have been raised about the last of these approaches, concerning ‘the anachronism of long data series’ (Weber, 1992, pp. 90–119). ‘To disassociate an object from its construction, at least theoretically, is to fail to recognize that measurements of certain objects depend totally on conventions of definition and encoding’ (Desrosières, 1992, p. 93). The hypothesis of this article is that the general increase in the activity rate recorded beginning with the 1896 census resulted from the implementation of the census’s new objectives, involving not only procedural improvements but also changes in classification criteria.
3 Unlike the genesis of the category of the ‘unemployed’ (Salais et al., 1986; Topalov, 1994; Zimmermann, 2001; Reynaud, 2018; Lagneau-Ymonet and Reynaud, 2020), the construction of the partition of the population into ‘active’ and ‘inactive’ individuals has been relatively little studied, despite the importance of this binary in economic and social policy. Existing studies on the topic have broken down its historical evolution into classification problems and practical problems, without addressing the administrative, political, or social objectives to which they respond (Desrosières, 1977; Topalov, 1999; Fouquet, 2004). Studying the genesis of an instrument requires knowledge of the objectives that governed its construction; but the archives used thus far to study the formation of statistical categories of labour (in particular, the published volumes of census results) make this exercise in reconstruction a difficult one (Desrosières, 1987). To overcome this obstacle, this article draws on sources that have been little used to study the construction of the active/inactive partition in France: the archives on the organization of the census at the Centre for Economic and Financial Archives and the Journal de la Société de statistique de Paris (Box 1).
4 In investigating the processes that led to the construction of the partition of the population into active and inactive individuals in the second half of the 19th century in France, this article follows in the line of work over the last several decades on the history of quantification (Mackenzie, 1991; Porter, 1995; Hacking, 1999; Desrosières and Thévenot, 2002 , 2008, 2010 ). This body of research is structured by three main ideas. First, the development of statistics is based on objects that cannot be immediately captured; they are the product of ‘conventional work’ (Desrosières, 2010 , p. 7). The categories and tools constructed and mobilized by administrations are not pure products of statistical reason, and they tend to depend on power relations and struggles around their development (Bourdieu, 2015). These constructions present themselves both as the ‘reflection’ of a context and as ‘agents’ in its production (Escafré-Dublet et al., 2018). They reconfigure society by creating ‘a new way of thinking, representing, and expressing the world, and acting upon it’ (Desrosières, 2008, p. 11). The act of classification is thus a political act (Schor, 2009). Finally, the validity of the constructed classifications and nomenclatures rests on the authority of the actor or institution that produces them (Porter, 1995; Bourdieu, 2015). The statistical administrations established by states during the 19th century enjoyed a double legitimacy linked to the authority of the State and of science (Desrosières, 2008).
Box 1. Selection of sources
The archives concerning the organization of the censuses used here are held at the Centre for Economic and Financial Archives (CAEF). They are available for each year starting from the 1896 census. Each collection includes a diversity of documents: not only official archives (committee reports, circulars, etc.) but also archives internal to the census department (reports and letters from census departments to the Ministry of Commerce, funding requests, instructions to counting officers, and documentation on remuneration and recruitment exams). The collection on the positivists recruited by the Labour Office, held by the Maison Auguste Comte (in particular the correspondence of Isidore Finance), offers some insights into how this administration functioned, but it does not provide additional information on the partition, particularly given that I had access to letters and reports sent by the directors of the administration to the Ministry of Commerce held in the CAEF archives.
To cross-reference instructions concerning encoding practices and the objectives of the partition, on the one hand, and the partition as it was ultimately constructed at the aggregate level, on the other hand, I draw on the tables in the Annuaires statistiques de la France (Statistical Yearbooks of France), also featured in the published census volumes.
Finally, because the census was not centralized until 1896, accessing archives related to its organization before that date is considerably more difficult. However, some documents from the archival collections used here do extend back to the organization of pre-1896 censuses; I crosschecked this information with publications in the JSSP.
5 In emphasizing the importance of the innovations that motivated such a construction or made it possible, this article also intersects with studies systematically exploring important aspects of the rise of statistics: whether studying the procedures for classifying a specific population and the resulting debates (Schor, 2009) or analysing the institutions developed to conduct population science in the 19th century and their legitimacy (Schweber, 2006). It is intended to contribute to the historiography of French public statistics and to contemporary debates on the quantification and evaluation of public policies, particularly those on the overhaul of the nomenclature of occupations and socio-occupational categories (Amossé, 2012). Some of the criteria used to define these categories today also played a role in the two partitions of the population into active and inactive subpopulations in late 19th-century France.
6 The article begins by surveying the partition of the population into active and inactive subpopulations in French censuses from 1861 to 1891 (Section I). For several reasons, a critique of the existing construction of occupational classification, along with a drive to produce a general census of the ‘productive forces’, arose in the 1880s (Section II). This census was conducted in 1896 and required several innovations (Section III). The foundations of this new partition responded to the objectives set for the census by the nascent labour administration, and are similar to those used today (Section IV).
I – Dependence on the household scale as a partition criterion (1861–1891)
7 In the census of 1836, a column for occupation appeared in the census lists, which at the time represented the main document available to count the population (Biraben, 1963).  For each household, the household head was listed first, followed by his wife, his children, other members of the household, and finally any servants living in the household. The occupation of the head of household was often the only one that was provided and then attributed to all members of the household. But this affiliation does not seem to have followed any specific rule: ‘most often, either the occupation of the head of the household was indiscriminately attributed to his spouse and even his children, or, for persons with two occupations, one of the two was not recorded, although we do not know which’ (Le Mée, 1979, p. 266).
8 From the second half of the 19th century, these lists of names constituted no more than a ‘second-hand’ document for the census. The use of household enumeration forms became national policy in 1856, and the use of individual forms in 1872 (Biraben, 1963). In the 1856 census, the objective of the occupational classification was to provide information on the number of individuals ‘directly’ and ‘indirectly’ dependent on each occupation. A circular issued by Statistique générale de la France (SGF)  stipulated that the census of 1856 was to ‘classify, within each occupation, not only the head of the household, but also all persons for whom his occupation provides directly or indirectly — his family, his workers, his employees (agents) of various kinds and even his servants’ (Le Mée, 1979, p. 273). Individuals practising an occupation and those they ‘provide for’ were thus systematically ‘confounded in a single number’ (Cahen, 1953, p. 231). This construction was not precisely repeated in the following census. Beyond providing information on the number of individuals directly and indirectly dependent on each occupation, the classification used in the census was to differentiate individuals directly practising an occupation from those ‘living on the income earned by the former’ (Loua, 1880). This division took shape in the 1861 census. The two groups were later named the ‘active population’ and the ‘inactive population’ for the 1876 census by Toussaint Loua, then director of SGF (Loua, 1879b).
9 Between 1861 and 1891, the active population was composed of individuals ‘directly’ practising an occupation in a particular position: business owners or employers; assistants or employees; labourers; or day labourers, cleaners, etc. It included household heads, but also other household members who reported practising a distinct occupation. The inactive population was constituted of ‘family members and domestic workers living from the work of the aforementioned’ (Ministère de l’Agriculture et du commerce, 1879, pp. 40–41).  These two groups represented the entire ‘classified’ population. The latter was thus divided both by occupations, themselves grouped into ‘occupational divisions’ (i.e. sectors of activity), and by position. Figure 1 shows the composition of the classified population. The total numbers in the separately counted population  and the ‘unclassified’ population must be added to the total number in the classified population to derive the size of the general population. The unclassified population included beggars, vagrants, registered sex workers, and individuals whose occupation was unknown.
Figure 1. Construction of the division of the population into active and inactive subpopulations (1861–1891)
Figure 1. Construction of the division of the population into active and inactive subpopulations (1861–1891)
10 Inactive individuals were counted under the total for the occupation of the individual at whose ‘expense’ they lived. For example, members of a household whose head was a civil servant and who did not report practising a distinct occupation were counted as inactive but systematically attached to the occupation of civil servant within the category of ‘the professions’. The first partition thus measured the relative size of different branches of activity not in terms of the number of individuals employed in each one, but of the number of individuals directly and indirectly dependent on those individuals:
The items prepared … by the administration were designed to answer two questions: How many individuals of either sex directly practise a given occupation, whether as heads of an establishment, employers, or as employees or labourers? How many individuals of each sex … are dependents of the former?
12 The partition thus captured the composition of the active and inactive population by position, and more curiously, because of the way it was constructed, the share of the total classified population (i.e. the aggregate of the active and the inactive) in each occupation and occupational division. This observation allows us to better understand why rentiers, in the occupational division for persons living exclusively from their income, were classified as active in the 1861 census and in the six following censuses.  Servants were considered inactive, as they were fed and housed within the household, like family members not practising an occupation distinct from that of the household head. This classification probably has to do with the nature of their activity; for example, since the work of domestic workers ‘in personal service’ was similar to that performed by other individuals (especially women), it was certainly perceived as non-productive. In the construction that prevailed between 1861 and 1891, servants were not considered to have an occupation as such. Their status was linked to their engagement by a single household. The classification of domestic workers as part of the inactive population was linked to this particular status; each year, they received a payment called a gage [as in the English word engage], which did not constitute remuneration for their services but a quantity (of money, food, etc.) paid ‘as a consequence of [their] engagement’  and which provided for their upkeep in the household. This same distinction existed under the Ancien Régime: the gage was the sum paid to servants, while wages constituted payment for the performance of specific tasks (Sarti, 2019).
13 The criterion of dependence upon which the partition was based between 1861 and 1891 consisted not only in relations of economic dependence but also in relations of legal dependence on the head of household. Within this construction, populations considered to be ‘minors’ were often associated with the inactive population and connected to the individual at whose ‘expense’ they lived. This suggests that characterization as ‘active’ or ‘inactive’ was derived from a social status and that actual participation in the production system was secondary, which can be seen in the very construction of the partition. It provided information on the distribution of the population by social position, not occupation. Therefore, in the construction that prevailed from 1861 to 1891, it was not participation in the production system that was captured, but the ability to live directly from one’s own income, regardless of whether it was linked to employment. In 1896, the criterion changed: rentiers were now excluded from the active population, while domestic workers were included. The instructions for the officers charged with counting the 1896 census data emphasized that domestic servants should be considered active, whereas the forms of the women ‘who reported being housekeepers’ but who only worked ‘in their own household’ were to be ‘rejected and set aside’.
14 Examination of the publications on the results of the 1861–1891 censuses, mainly produced by Toussaint Loua, suggests that the goal of the earlier partition was to construct ‘social classes’ (Loua, 1873, 1874a, 1874b, 1879a, 1879b) based on the position of the household head. This construction allowed Loua to establish demographic characteristics for each class, analysing differences in fertility and mortality rates at the household level. A second objective seems to have been to monitor the population growth of the different classes (Loua, 1874a), again comparing their fertility and mortality rates. Behind the earlier partition of the population were thus two main objectives. The first was to organize the structure of society into different social classes marked by a double dependency relationship: within the household and between employees and business/property owners. As Cahen pointed out, until 1891 the goals of the occupational censuses were ‘primarily sociological’ (Cahen, 1953, p. 232). The second objective was to establish the demographic characteristics of each class, enabling analysis of their growth rates, among other things. By ranking the population by social class and sketching the relations of dependence connecting them to each other, statistics on occupations and positions developed during this period seem to have been a part of the movement of mise en équivalence (making equivalent) of individuals, allowing the ‘masses’ to be thought (Desrosières, 1988). Analysing the progression of different classes, then, could offer a means of monitoring the social order’s sustainability by enabling the anticipation of future imbalances linked to the faster growth of one class compared to another, as Loua suggested (1874a). Moreover, his writings also show that the core aim of the classification of occupations and positions was above all to represent the social order, not the structure of the available workforce: ‘social functions’ were distinguished by ‘hierarchical order’ (Loua, 1874b), with rentiers and business/property owners at the top.
15 The choice of the household scale for the 1861–1891 partition also constituted a reflection of the predominant mode of organization of labour, at least for some time. Large organizations were still rare, and trade and production remained widely dispersed. Individuals working at home were central in this mode of organization, performing their activity without direct supervision ‘except within their household’ (François and Lemercier, 2021, p. 16). Beginning in the 1880s, the reliability of the occupational classification constructed in the context of these censuses was gradually challenged and ultimately found inadequate, with concerns linked to the rise of industry and the development of unions. The establishment of a census of the ‘productive forces’ progressively came to be seen as necessary for implementing insurance laws related to work.
II – The turning point of the 1890s: context and motivations for a census of France’s ‘productive forces’
16 With the 1881 census—the first to count the population of France as a whole on a fixed date—a new category was added to the occupational divisions, distinct from that of individuals with an unknown profession: ‘individuals with no occupation’. In the results of the 1886 and 1891 censuses, this category was then combined with that of ‘unclassified or unknown occupations’. It included street performers, bohemians, vagabonds, and sex workers, as well as people with no position (gens sans place, an expression used to refer to unemployed individuals), and, in its connection to the unclassified population, was associated with ‘children confided to a wet nurse, students, or pupils in boarding or residential schools, resident personnel of asylums, hospitals, hospices, etc., foundlings’. The history of shifting ways of categorizing individuals with no occupation is interesting on multiple grounds. On the one hand, they prefigure hesitations in the construction of the partition concerning individuals with no occupation (how would unclassified individuals be distinguished from individuals with no occupation?) as well as gens sans place (how to distinguish between individuals with no occupation and the unemployed?). These questions would be partially settled in the occupational census of 1896. On the other hand, they reflect a concern, which grew from the 1880s onward, with what François and Lemercier (2021) called the ‘age of the factory’: the gradual disappearance of the workshop in favour of the factory, the emergence of salaried employment as a reference, and involuntary unemployment, accentuated by the Long Depression that would strike France in the 1880s and by the phylloxera crisis and the resulting acceleration of the rural exodus. Finally, the introduction of the distinction between the ‘unemployed’ and ‘individuals with no occupation’ reflects a major shift in the long-standing tension between the understanding of indigence as a social phenomenon and the notion of individual responsibility. In this context, individuals gradually came to be thought of as subject to ‘risks’ linked to industrialization for which they could not be held responsible, particularly if unemployed (Ewald, 1986; Rosanvallon, 1990).
17 From the early 1880s, a growing range of statistics were compiled on wages in industry, as well as on unions and strikes—developments not unrelated to the law of 1884 legalizing unions and various occupational organizations. Major parliamentary inquiries were conducted at the request of the Chamber of Deputies, such as the 1884 Survey of the Situation of Workers in Agriculture and Industry in France, and on the Crisis of Paris. Finally, the Higher Council of Labour and the Labour Office were created in 1891. On its creation, the Labour Office was made a department of the Ministry of Commerce, under the authority of the Minister. Because the Ministry recognized statistics and surveys as a major tool for action and knowledge, and given the broad scope of the missions entrusted to it, the Labour Office was granted considerable freedom of operation. It was conceived as a statistical office whose purpose would be to collect, coordinate, and publish all statistics on work and employment. It was also tasked with providing input that would inform the development of labour legislation: this conception of the Labour Office can be seen in the first sessions of the Higher Council of Labour, as well as in the declarations of its various directors. In 1893, Jules Lax, the director of the Labour Office from 1891 to 1893, wrote the following to the Minister of Commerce, Louis Terrier, in a letter presenting the first volume of the survey of wages in French industry:
To preface all our studies with a general survey of the productive forces of France, animate and inanimate, would doubtless have been more consistent with the rules of pure logic. It is evident indeed that until these essential factors in our industrial activity have been clearly identified, most of the laws concerning industry and workers—which public opinion demands and which the government of the Republic aspires just as ardently to provide—could be based only on more or less bold conjectures and more or less fortunate intuitions. … To what grave miscalculations might we expose ourselves, for example, were we to enact sickness or old-age insurance laws based solely on knowledge of wages, working hours, and even occupational risks, but in the absence of definite data on the absolute size of the workforce, or at least the relative size of the various categories of workers?
19 The project of an occupational census thus reflected the ambition of applying a scientific method to the implementation of insurance laws around work, intended to regulate various ‘occupational risks’ (unemployment, workplace accidents, illness, and old age). Like Émile Cheysson, who sat on the standing committee of the Congrès internationaux des accidents du travail (international congresses on workplace accidents) and argued that an occupational census would provide a quantitative response to the associated debates, Jules Lax—followed by Arthur Fontaine, Director of Labour from 1899 to 1919 (Box 2)—linked the need for an occupational census to the various proposed insurance laws. Until the early 1880s, the view of statistics as a tool to be used ad hoc to respond to conjunctural problems, as seen for example in the role given to parliamentary inquiries, seemed to push the census data on occupations into the background. The weaknesses of the existing classification were pointed out as early as the 1870s, but more systematic critiques were only developed in the mid-1880s. The growing interest in census data was part of a wider movement in favour of the institutionalization of the census, as reflected by the increasingly general use of new census documents, the setting of a fixed date for the census (1881), and efforts to define the population that the census should cover and particular categories of population (such as the separately counted population). Other developments that reflected this process of institutionalization include the creation, in 1885, of the International Statistical Institute and the Higher Council on Statistics, as well as the quest to establish a unified scheme of occupational categories to allow comparisons over time and between countries. The International Statistical Institute charged the French statistician and demographer Jacques Bertillon with the task of developing a nomenclature of occupations for use in the 1896 census.
20 The turning point in the justification for the census advanced by the nascent labour administration system can be understood at different levels. At the administrative level, the census may have been strategic for developing the Labour Office’s activities. The Labour Office was a small structure, employing 15 to 30 people between 1891 and 1914 (Lespinet-Moret, 2007). The integration of the census into the Labour Office’s programme, through the creation of a special department and the affiliation of SGF, broadened its prerogatives while strengthening the technical and political justification of the census that had been in construction since the mid-1880s. At the national and international levels, the Labour Office directors’ declarations on the importance of an occupational census can also be understood as endogenous: the creation of a law on workplace accidents was still in debate in the early 1890s, and the commission charged with developing the plan for an occupational census in France referred to the German census of 1882. The commission’s report features the idea of an occupational census as a basis for future insurance laws, which may have influenced the objectives given to the 1896 French census:
It [the commission] had to look abroad for examples. … The German Empire, whose example had already been cited in Cheysson’s report, appeared to the commission to have achieved the widest coverage of the subject and to have best achieved the intended goal with the occupational census of 1882. This operation served as the basis for the establishment of laws on workers’ insurance. It is interesting to add that the German government seems to be considering making this type of enumeration periodic, which it deems necessary for its economic policy: it is preparing to continuously revise that policy using the same procedures and the same questionnaires, slightly modified.
Box 2. Biographical notes
Émile Cheysson was an engineer in the Corps des ponts et chaussées (Corps of Bridges and Roads). He was a member of the Statistical Society of Paris and its president in 1883. He contributed to the creation of the Higher Council on Statistics (Desrosières, 2008).
Arthur Fontaine, who graduated from the École Polytechnique in the same year as Lucien March, was appointed deputy director of the Labour Office in 1894 and became Director of Labour in 1899 on the proposal of Alexandre Millerand. He held this position until 1919. He was also the first president of the International Labour Organization.
Jules Lax was an engineer who studied at the École Polytechnique (1862) and the École des Ponts et Chaussées (1864). He was appointed director of the Labour Office on its creation (1891), while he was Inspecteur général des ponts et chaussées. He held this position until 1893.
Émile Levasseur was a statistician and geographer who studied history at the École Normale Supérieure. He was a member of the Higher Council on Statistics and chaired the Statistical Society of Paris in 1877 and 1900. He also participated in the International Statistical Congresses and became vice-president of the International Statistical Institute in 1886 (see Palsky, 2006).
Toussaint Loua was the director of SGF from 1875 to 1887. He began his career as a clerical employee at the Ministry of Commerce in 1850 at age 26. He obtained his first post at SGF in 1853 and worked there until his retirement. He became a permanent member of the Statistical Society of Paris in 1864, and participated in the creation in 1885 of the Higher Council on Statistics, of which he was also a member.
Lucien March, a graduate of the École Polytechnique, became permanent delegate to the Labour Office in 1892. He joined the Statistical Society of Paris in 1897 and became its president in 1907. After directing the counting department for the 1896 census, he was appointed head of technical services for the Labour Office and SGF in 1901 by Arthur Fontaine. This promotion followed the incorporation of the Labour Office into the Directorate of Labour. March became director of SGF in 1910, a position he held for 10 years. He was the first director of the SGF with scientific training (Kang, 1992).
22 In the specific case of unemployment insurance, Germany and France both counted the unemployed at the national level before implementing an unemployment benefits system, unlike Great Britain, which relied on figures provided by unions (Reynaud, 2018). It seems that the census’s justification advanced by the labour administration required the construction of a representation of activity and occupations that was ‘functional’ for legislators and the public authorities, and thus attuned to those key social problems of the time: ‘occupational risks’ such as workplace accidents or unemployment. For the constructed representation of activity to be functional, it had to be directly mobilizable, operationalizable, and thus ‘action-oriented’ (Perrot, 1973, p. 15). That the 1896 occupational census forms for inactive individuals were not counted, and that no comment on the structure of this population was included in the published census report, is itself an important characteristic of the newly constructed partition. Individuals without an occupation were excluded by its construction from social legislation; or, conversely, the partition was the product of their exclusion from political discourse. The ‘social question’, a notion often evoked at the time by labour administrators, seems to have been polarized around the protection of workers, especially those in industry. While the partition constructed in the pre-1896 censuses supposed a relationship of dependence between the active and inactive populations, with the 1896 census its primary objective was to distinguish the productive population—the active—from the unproductive remainder of the total population—the inactive.
III – The 1896 census: starting point for the ‘industrialization’ of statistics
23 Until the 1891 census, local prefects were tasked with centralizing the documents filled in by the municipalities and producing the summary tables for each one. They did not always have the means to check the quality and veracity of the information transmitted to them. SGF then took charge of assembling the documents and drawing up the tables for the mid-level administrative divisions (départements) and for France as a whole. The lack of tools for checking census data and the weaknesses of the classification system used in previous censuses (Loua, 1880, 1888), coupled with the desire to establish national statistics on occupations, led to the creation of a commission responsible for drawing up the plan for a general occupational census. The commission was composed of around 10 members, including Émile Levasseur, Jacques Bertillon, Émile Cheysson, Arthur Fontaine, and Victor Turquan (then director of SGF) (Box 2). The report Levasseur presented to the Minister of Commerce proposed the option of an occupational census annexed to the population census because of the costs a special census could generate and the possible mistrust of the population towards an additional census. Although the report presented this option as a less effective one, it was adopted for the 1896 census.
24 The partition constructed in this context reflected the design of the individual forms, which resulted from the directions of the commission, as well as the work of the counting officers required to set aside all the forms for individuals with no occupation that reached the department by mistake.
25 With the attachment of the occupational census to the general census, the individual census form was separated into two distinct parts: one for the general population census and another with questions on occupation and position. A census officer distributed the individual and household forms to each household. After verification, a file containing a summary and both forms was sent to the local mayor. The municipal administration detached the occupational form, which then had to be allocated to one of four categories: individuals with no occupation, separately counted population, individuals who have an occupation but are currently unemployed, and individuals currently practising an occupation, grouped by the address of the establishment where they were employed. An ‘establishment’ was defined in the instructions for the 1896 census:
The combination of several individuals habitually working together, in a specific house or building, or in multiple neighbouring houses or buildings, under the direction of one or more representatives of a single company. … An individual working alone at home is considered to constitute a separate establishment.
27 These packages were then sent to the prefecture, which, after verification, forwarded them to the head office of the Labour Office to be counted, except for the forms of individuals with no occupation. These last were not counted, but kept by the prefecture until permission to destroy them was granted. 
28 The major innovation in the occupational census of 1896 lay in the counting process. While the results from the first part of the form continued to be counted at the municipal level, the counting of the occupational part was centralized in Paris. It was performed by a department directed by Lucien March, using Herman Hollerith’s punch card machines, which had already been used in the US census and the Austrian census of 1890. This department employed around 100 people between 1896 and 1900 (Huber, 1937). On receiving the occupational forms, the department’s counting officers had to perform a series of preparatory steps ahead of the machines’ information processing operations. The first was to determine the nature of the industry to which each establishment belonged and then to punch it into the top-ranking individual form in the set for a given establishment. After correcting the number of persons employed in the establishment if needed, the counting officer checked the information declared on each form, referring to instructions for forms that could not be immediately used because they were incomplete or falsified. Some forms were set aside at this stage by the officers and were not subject to further processing. This occurred, for example, with the forms of individuals who reported practising an occupation but were subsequently classified as inactive by the counting department (e.g. women who reported being ‘housekeepers’ but whose activity was restricted to their own household). Next, the forms were identified by industry number. They were assigned a serial number, also written on a card punched according to the information on the form. Finally, these cards were classified by sex and perforated by arrondissement and département. They were then ready to be read by the Hollerith machine. All the information reported on the form by the individual was encoded on the card by numbers or combinations of letters. For each piece of information indicated on the card, a hole was read as an affirmative response, and the counter for this entry was then increased by an increment of 1. If no hole was punched for a given entry, the count remained unchanged. The items entered via the card were divided into series and could be combined, notably to produce tables with multiple entries. Verification mechanisms were integrated into the machine (audible signals, displays of totals) to alert users to any errors (Cheysson, 1892).
29 The use of Hollerith machines in the 1896 census represents an important stage in the history of statistics. It constitutes the starting point for the automatic processing of census data and thus for the ‘industrialization of statistics’ (Desrosières, 2008, p. 272). In this respect, it represents a departure from the methods previously used to process census data, whose counting was still largely decentralized in 1891. However, the use of these new methods also took its place within a longer process of institutionalization of the census and improvement of the methods used to enumerate and study the population, both at the national level (use of household and individual forms, conduct of the census on a fixed date, creation of the Higher Council on Statistics, etc.) and the international level (creation of the International Statistical Institute, the quest for an international nomenclature of occupations). Finally, the use of the Hollerith machines represented a condition for the establishment and processing of the occupational census, which at the time involved a considerable volume of work. While their use accelerated the processing of the occupational forms from the 1896 census, the process nonetheless took 4 years, whereas the commission had estimated in 1894 that it would take at most 3.  The first results of the census (for about 15 départements) were published in 1899, while the results for all départements and for France as a whole were included in the Statistical Yearbook of 1900, only a year before the following census. The time savings with the new methods were nonetheless very substantial: in comparison, the results of SGF’s industrial survey of 1860–1865 were not published until 1873, 13 years after the process had begun (Chanut et al., 2000).
IV – Market labour as the foundation of the active/inactive partition
30 The partition constructed from the 1896 census was founded on the distinction between the unemployed and individuals with no occupation. Individuals who reported having an occupation were considered active, whether they were employed or unemployed at the time, while all individuals with no occupation were classified as inactive. This construction eliminated the criterion of dependence at the household level, highlighting instead the individual’s dependence on the establishment that employed them—the objective having been to establish statistics at the establishment level as well.
31 The last section of the individual form featured three items: the first concerned the individual’s occupation, while the other two related to their position within it. The first was used to distinguish individuals with an occupation from those without one. The other two items aimed ‘to connect persons practising an occupation to the company that currently provides them with the means of existence’ (Direction générale de l’INSEE, 1896 census, Instruction of 10 February 1896 on operations for the enumeration of the population). Individuals could practise their occupation either inside or outside their household. In the definition of an ‘establishment’ used for the 1896 census, the family unit was an establishment, and the household head was also the head of an establishment. Instructions on the back of the form indicated that individuals who worked at home ‘under no one’s supervision’ were to consider themselves heads of an establishment and indicate the number of assistants they employed, ‘even from their family’. Family members were to indicate the occupation they practised under the supervision of the head of household.
32 Individuals without a position on census day were considered unemployed if they usually practised an occupation as employees or workers, under the supervision or in the service of others, in a relationship necessarily involving affiliation to an establishment.  Three defining criteria were applied to distinguish individuals without an occupation from the unemployed: unemployment was temporary, linked to the suspension of work in an establishment, and only applied to individuals habitually practising an occupation. Legal dependence was the key criterion for the individual’s affiliation to the establishment that employed them (Salais et al., 1986): only individuals physically working within an establishment were to be counted by the heads of establishments; homeworkers, even if they depended economically on an establishment, were not linked to it in the census. Regarding the distinguishing criterion of dependence highlighted above for the previous partition, then, several changes had taken place: the relationship of dependence at the household level was lost, and the relationship between ‘social classes’ gave way to the individual’s relation of dependence with an establishment.
33 The structure of the table in the 1900 Statistical Yearbook presenting the composition of the working population of France as a whole is similar to those of previous years. It is broken down into ‘industries or occupations’, themselves divided into subgroups, and the positions of individuals. The active population included heads of establishments, the employees and workers of establishments, and ‘dispersed workers, small business owners, and outworkers or workers without a single fixed position’ (Ministère du Commerce, de l’industrie, des postes et des télégraphes, 1900, p. 22).  Additionally, it included ‘persons in an unknown situation’ and ‘employees and workers with no position’. The category of the unemployed was thus included in the active population, unlike the category of individuals with no occupation. This distinction was linked to the new objective of the occupational census: to represent the totality of the country’s productive forces. The unemployed were considered active because they usually had an occupation and were expected to quickly find another job; in other words, because they were available on the labour market. Forms from ‘individuals with no occupation’—who answered ‘none’ to the question ‘What is your occupation?’—were parcelled together and kept by the prefecture until permission to destroy them was granted. They were simply not included at all in the counting of the occupational census data. The inactive population consisted of all individuals with no occupation, whose forms were not processed beyond this initial classification, along with individuals whose forms were set aside during the counting officers’ subsequent processing work. Thus, the forms of the individuals constituting the inactive population were not processed further: they were simply separated from the rest of the forms. No table in the Statistical Yearbooks shows the composition of the inactive population over the period 1896–1936, although the forms of the inactive were counted starting in 1901.
34 The change in the partition criteria led to several noteworthy changes. First, the activity rate considerably and durably increased, with a marked rise for women (Table 1). Between 1876 and 1891, the proportion of women in the active population fluctuated in a narrow range around 30%; in the 1896 census, women represented 34.6% of the active population, an increase of 5.1 percentage points from 1891 to 1896. From the beginning of the 20th century, women on the labour market were more consistently counted than during the 19th century. This has been attributed to the implementation of new counting methods and the centralization of the counting process (Maruani and Meron, 2012). Several likely reasons may explain why the proportion of women counted as active increased with the changes in the partition. The main cause of this increase is the shift from the household scale to the individual scale. This eliminated the dependence criterion within the household: classification as active or inactive no longer depended on the individual’s position within the household, but on their individual practice of an occupation, in the sense of participation in a labour market.  As noted above, this labour could be performed either within or outside the household. Consequently, beginning in 1896, many women previously classified as inactive because they worked within the family structure were counted in the active population. A condition of this ‘fragmentation’ of the domestic sphere (Topalov, 1999, p. 402) via the erasure of the dependence relationship within the household, however, was the delimitation of what would count as productive activity, particularly within the family unit. Here, the distinction established in the 1896 census between ‘domestics’ and ‘housewives’ represents an indicator of the emergence of the definition of activity as participation in market exchange. For an individual to be active was thus for them to work ‘in the limited field of production’ (Fouquet, 2004, p. 55).
Table 1. Active population in France (number and proportion) by sex, 1876–1906
Table 1. Active population in France (number and proportion) by sex, 1876–1906Note: Dashes indicate that the category of individuals of unknown sex did not appear in the corresponding statistical tables.
35 A second significant change should also be noted. From 1896 onwards, the occupational divisions, which had previously included both active and inactive individuals, consisted only of active individuals. The totals presented for each occupation and occupational division thus reflected the number of active individuals whether employed or unemployed, and not the aggregate total of the active and inactive population. The distinction between the unemployed and individuals with no occupation yielded this new partition by instituting the individual’s capacity to sell their labour on a market as a distinguishing criterion. With this partition, the inactive population became a heterogeneous residual category, of unspecified composition. It ceased to have its own definition and ‘resulted at best from a double negation: those neither active in employment, nor unemployed’ (Fouquet, 2004, p. 47). This residual character was the product of a context marked by a polarization of discourse around work: the aim of the insurance laws that were being prepared at the end of the 19th century, which were linked from the mid-1880s to the need for a census of productive forces, was to regulate certain ‘occupational risks’. The lack of interest in the inactive population reflected in the construction of the partition is thus ‘consistent’ (Didier, 2009) with the related objective of the census: these laws were aimed at the individuals who made up the active population.
36 The changing partition of the population into the active and the inactive in late 19th-century France reflected a combination of technical, theoretical, and political choices. The first partition, used for the censuses of 1861–1891, mainly reflected sociological and moral considerations. It motivated an analysis in terms of social class, providing the instruments to analyse change over time in the size of different social groups, hierarchically classified, by quantifying their mortality and fertility. The objective behind the later partition, applied from 1896 and similar to the one used today, was linked to the nascent system of labour administration: to represent the state of the country’s productive forces. While the partition constructed for the censuses of 1861–1891 was little used outside the institution that produced it, the one applied from 1896 onward would be an instrument for administrators, legislators, and the public authorities.
37 The pursuit of this new objective also led to the application of new procedures. Changes in the partition thus at once reflected and drove transformations of labour during this period. It reflected such transformations because it manifested multiple important shifts in the representation of activity: from the household scale to the individual scale, from family activity to collective activity within establishments, and from social position to contribution to the productive system. It also drove such transformations because it paved the way for the use of employment statistics as a policy instrument. However, in the late 19th century, the relationship between the development of a national statistical apparatus and the creation of labour laws was not totally explicit. The law of 1898 on workplace accidents, for example, was passed well before the counting of the 1896 census had been completed.
38 The two competing logics in these different partitions can be observed in the nomenclature of socio-occupational categories used today. According to Alain Desrosières (Desrosières and Thévenot, 2002 ), the construction of this nomenclature can be understood in three phases: organization by trades (which prevailed under the Ancien Régime); the distinction between wage- earning workers (salariat) and others (non-salariat) from the 1850s; and the introduction of a ‘hierarchy of wage earners’ based on training after the 1930s. Between 1896 and 1936, ‘the information on occupations and “situations within a profession/occupation” provided by the censuses was mainly intended to describe the socio-economic characteristics of production and the workforce, rather than more or less hierarchical social status’ (Desrosières and Thévenot, 2002 , p. 8). Along with the emergence of the distinction between wage-earning workers and others between the 1850s and 1896, census statistics on occupations and positions effectively described hierarchical social statuses, which were not necessarily classified by skills or know-how but based on social, economic, and legal criteria. This can be seen in the criteria used in the partition of the population applied in the censuses of 1861–1891. The nomenclature of socio-occupational categories is thus based on the synthesis of several criteria, traces of which can be found in the construction of the partition between the active and the inactive: division into different social classes, hierarchical classification by qualifications and skills, and division by occupation and social position.
On the reliability of the results of general censuses in France in the 19th century, see Le Mée (1979).
A Bureau de statistique générale (general statistics office) was created in France in 1833 on the initiative of Adolphe Thiers. It was renamed as Statistique générale de la France in 1840. For its history, see Huber (1937).
This population consisted of individuals who did not fall under the scope of municipal laws or tax: for example, individuals residing in workhouses or in public high schools and colleges.
From 1801 to 1946, the census took place every 5 years in France, with a few exceptions: the census of 1816 did not take place; the planned census of 1871 was conducted in 1872 (it was delayed by the Franco-Prussian War); and finally, those of 1916 and 1941 were not carried out due to wars. Until the 1881 census, census officers filled in the census documents themselves, which led to long delays: 6 to 8 weeks, for example, in 1876 (Biraben, 1963). Beginning in 1881, the census was conducted at the national level on a fixed date; officers would drop off the forms in advance to be filled in by the household, and collect them on the census day.
Littré É., 1873–1874, Gage, in Dictionnaire de la langue française. Retrieved 19 September 2020, from https://www.littre.org/definition/gage
The occupational form for the 1896 census requested no additional information from individuals who indicated that they had no occupation. The code ‘SP’ (for ‘sans profession’, without occupation) was carried over to the first part of the general census form.
This additional 1-year delay may have been due to the unforeseen proportion of forms classified as incomplete during the counting process, which led to the hiring of additional officers.
In addition to the temporary nature of unemployment, on the 1896 forms it was identified through its causes (illness or disability, regular off-season, work shortage due to other accidental factors).
INSEE defines ‘occupational activity’ as ‘work performed with a view to a commercial/market exchange … and whose nature is not contrary to the law or to public morals’.