1Confronted with the critique of sources such as population censuses, students of family labour, family income and women’s work have turned to another and allegedly “better” source: budgets of working-class families’ income and spending, compiled from the beginning of the 19th century onwards2. The shortcomings of family budgets did not pass unnoticed. Their focus on better-off working-class families, their preference for nuclear families and the questionable reliability of data on income and expenditure have all been acknowledged. However, after the “inevitable” criticism of the source, social and economic historians tend to rush for “the figures”. The study of the Belgian budget inquiries of 1891 and 1928-1929 suggests, however, that these historical statistics teach us more about the context in which they were executed than about the object of their inquiries. Acknowledging this implies a shifting research focus. The social, ideological and intellectual climate in which historical statistics were compiled and the administrative and practical “details” that governed their execution are promoted from “background information” to the status of a major research object.

2Peter Scholliers and I took apart the Belgian family budget inquiry of 18913. This investigation was indirectly used as a means of attenuating class antagonism. The industrial jacquerie of 1886 had made a strong impression on the Belgian élite and when tension rose in the course of 1890 as a result of increasing socialist mobilisation for universal male suffrage, an inquiry into the condition of the Belgian working class was ordered. The official reason for the investigation was the fact that France had denunciated its commercial treaty with Belgium4. The inquiry was supposed to answer the question whether the Belgian working class was in a condition to struggle through an industrial crisis and possible wage reductions. However, in the parliamentary discussion that took place a month before the inquiry was ordered, the need for industrial and labour statistics was stressed without any reference to the commercial problem, although the latter subject had been on the agenda since the end of 1890. Instead, it was argued that in an era when some tried to agitate the working class in order to promote universal suffrage, the government might want to prove how devoted it was to working-class interests5. The organisation of the budget inquiry put to work the newly created organs of class conciliation (Conseils de l’Industrie et du Travail) where these were most needed, and its results reassured both reformers and the working-class élite that Belgian capitalism was on the right track.

5In the year following the end of the Great War, Belgium was characterised by strikes and social unrest. The loss of purchasing power as a result of galloping inflation was one of the grievances that triggered social agitation. In order to reduce social tension, the Belgian government, under the impulse of the socialist minister J. Wauters, introduced the principle of measuring inflation by compiling an index of retail prices6. In its issue of 1 May 1919 the Ministry of Labour’s periodical, the Revue du Travail, published its first price index7. In their negotiations, representatives of workers and employers could (but did not have to) use the price index to adapt wages to changing price levels. Conceived as a means of transforming potentially explosive class antagonism into negotiable matter, the index of retail prices itself, however, became the object of discussion. The inter-war period witnessed an endless debate as to how it should be calculated8.

6One of the most important points of criticism, expressed from the very beginning of its existence, concerned the fact that the index of retail prices was a non-weighted average of prices of products, regardless of their relative importance in consumption9. As early as 1920, the Ministry tried to counter criticism by publishing a weighted index of food prices, based on the results of an inquiry into food consumption held in 1910 by the Institut de Sociologie of the Université Libre de Bruxelles. The parallel development of this weighted index and the official index of retail prices was meant to reassure the interested parties that the latter was doing a fine job10.

7Assuming that consumption patterns had changed since 1910 and given the fact that only food expenditure had been included in the pre-war investigation, a new budget inquiry was called for in 192111. On the basis of the results of the inquiry into the budgets of 673 blue-collar workers’ families and 175 petty-bourgeois families, a weighted index of the cost of living was compiled. Again, however, the purpose of the exercise was not to replace the index of retail prices by a weighted variant. On the contrary, the parallel evolution of the two was meant to show that it was justified to continue to use the former. The government and social partners preferred to stick to the non-weighted price index because its evolution was more stable. In view of the existing controversy, however, the weighted cost of living index had to provide the straight average with a quality label. Both indices were published in the Revue du Travail12.

8In the light of widespread inflation, calculating cost of living indices was not limited to Belgium. Budget inquiries were organised in Denmark (1922), the United States (1918-1919), Japan (1919-1920), New Zealand (1919), Norway (1918-1919), the Netherlands (1917, 1918-1919), Finland (1920-1921), the Soviet Union (1922), the United Kingdom (1918), Ireland (1922), Sweden (1916-1918, 1920, 1923) and Switzerland (1921)13. In the second half of the 1920s and in the 1930s budget inquiries were conducted in some thirty countries14.

9The International Labour Office was a strong promoter of this kind of exercise. International comparison of the cost of living was one of its preoccupations. In this respect, it continued the comparative work initiated by the British Board of Trade round the turn of the century15. The second international conference of labour statisticians (1925) voted a resolution recommending countries to conduct a budget inquiry as soon as the post-war economic situation had improved and, if possible, the inquiry was to take place before the end of 192816.

10By 1928 these conditions had certainly been met in Belgium. The Belgian franc had been stabilised, the government had successfully tackled the problem of unsecured debt and the economy was booming. The recommendation of the conference of labour statisticians was probably the main reason for the organisation of the Belgian budget inquiry of 1928-192917. In contrast to what might have been expected, given the precedent of 1921, its outcome was not used for adapting the weightings of the cost of living index to changes in consumption patterns18.

11Neither the budget inquiry of 1921 nor the one of 1928-1929 therefore interfered directly in the emerging machinery of institutionalised industrial relations19. The weighted index of the cost of living, compiled on the basis of the inquiry of 1921 and meant to “correct” the index of retail prices, was never used as such. The main use of the inquiry of 1928-1929, so I will argue, was ideological. It demonstrated:

12There is no information available on the individual budgets gathered in the inquiry of 1928-1929, but we have two unpublished typewritten reports at our disposal dealing with its organisation (see further). They were found in the “Household budgets” section of the Belgian Institut National de Statistique. These internal reports received no publicity. For decades the only sources of information on the results of the inquiry were two short articles by Max Gottschalk and four contributions by Armand Julin, the éminence grise of Belgian statistics21. Julin’s most extensive article on the inquiry of 1928-1929 was published in the journal of the Institut International de Statistique22. He had presented this account during the London session of 1934, in the section dealing with social statistics. A summary of his presentation to the London conference was reproduced in the periodical Revue du Travail of the Ministère du Travail et de la Prévoyance Sociale23.

13The burden of the articles by both Gottschalk and Julin is clear: all six stress the tremendous improvement in the standard of living of Belgian workers. Max Gottschalk, a collaborator of the Institut de Sociologie of the Université Libre de Bruxelles, hardly commented on the fact that the inquiry had been conducted in a period of great economic prosperity and that since then things might have deteriorated24. In two of his articles Julin mentioned the current crisis without really going into the matter25. In a third contribution, partly reproduced in a fourth one, Armand Julin explicitly addressed the question of the effects of the economic crisis and of unemployment26. The improvement of the standard of living had perhaps been halted, he argued, but never again would the dark days of the 1840s return.

15The report that is available only in French is 26 pages long and was written by V. Lion, a collaborator of the Ministry of Economic Affairs. It was clearly written after the Second World War, since the commission and sub-commission mentioned in the title of the report were established only in 1945 and 1946 respectively27. It is much more difficult to date the more extensive French/Dutch reports, running to some 100 pages each. It is possible that they too were compiled after the Second World War (see Appendix 1). This does not imply, however, that the information they contain was previously unavailable. Quite a few tables in the Dutch/French reports are identical to those in Julin’s contribution28. Julin’s article, however, did not include the tables illustrating the methodological problems encountered by the organisers. Armand Julin himself is not very clear as to the material he used. In a footnote he refers to an Enquête sur les budgets de familles ouvrières et de la petite bourgeoisie 1928-1929, (inédit) without any further details29.

17For students of the 1891 budget inquiry, Armand Julin is an old acquaintance. Indeed, we meet him as its most authoritative commentator30. Educated at the University of Liège, where he studied law and political and administrative sciences, he was a pupil of Émile de Laveleye. The latter was a professor of political economy, Kathedersozialist and member of the Liberal party. Émile de Laveleye rejected laissez-faire but recognised the benefits of competition and was in favour of state intervention in the social field31. Armand Julin worked under de Laveleye’s guidance in 1889-1890 and his first publication dealt with wages in the Belgian mining industry between 1810 and 188932.

18In November 1890, Armand Julin joined the industrial directorate of the Ministère de l’Agriculture, de l’Industrie et des Travaux Publics, the administration that organised and published the budget inquiry of 189133. Despite its Leplaysian inspiration, the latter, as its name clearly states, inquired into the budgets of 188 working-class families and made no attempt to produce a series of Leplaysian monographs, which would have implied intensive observation of a very limited number of families.

19Belgian Leplaysians were not the only followers to propose a pragmatic adaptation of Le Play’s legacy. In France, the project of the so-called orthodox Leplaysians was quietly revised in the 1890s. The Leplaysian monograph was gradually replaced by statistics, and methods were no longer explored within the Société d’Économie Sociale, but in the newly created ministerial department, the Office du Travail, whose function was to inquire into the state of labour and the condition of the working class. In the Office du Travail, Leplaysians and the French state apparatus joined forces34. The Belgian Leplaysians followed the same track. They were well represented in the administration of industry and between 1884 and 1899 they provided the catholic ministers (the catholic party was in power from 1884 to 1914) who had industry and labour within their brief. The creation of a separate Ministère de l’Industrie et du Travail in 1895 was largely due to Leplaysian efforts35.

20A. Julin formed an almost archetypal example of the Leplaysian at the service of the state, exchanging the monograph for statistical calculations. In 1890, the year he joined the Ministère de l’Agriculture, de l’Industrie et des Travaux Publics, he was admitted as a member of the Leplaysian Société d’Économie Sociale36. In 1895, Julin produced his own Leplaysian monograph on a Liège metal worker and in the Leplaysian periodical La Réforme Sociale he published extensive discussions of the Belgian family budgets collected in 1891 and (rather haphazardly) in 188637. Gradually, however, he built up a reputation as a statistical expert. He directed statistical projects at the Department of Industry and Labour, published on the subject, and in the inter-war years he taught statistics at the universities of Ghent and Liège38. In a textbook on statistics from 1928, Armand Julin emphasised the difference between the Leplaysian monograph and statistics. In the former, he argued, mass observation and representativeness were lacking. Moreover Le Play’s ambition was to establish universal truths and this dogmatism ran counter to the statistical spirit39.

21In 1913, Armand Julin became head of the Office du Travail (the name was identical to that of its counterpart in France), a part of the Department of Industry and Labour that carried out its inquiries and which he had joined in 1895. He reached the highest administrative post of the Department in 1919, when he became its secretary general40. Despite the fact that he had become the administrative head of the Department, Julin was still involved in some of its statistical work. With regard to the statistics section of the directorate of labour, the Almanach Royal mentioned that “le secrétaire général dirige personnellement ce service”41. For Julin’s successors this was no longer the case.

22During the Great War, Armand Julin worked on a theoretical treatise of statistics, in which the calculation of indices occupied a prominent place. So it was to him that the socialist Minister J. Wauters turned for the calculation of the index of retail prices in 191942. The issue of index calculation would preoccupy Julin all through the inter-war period. Although he retired in May 1930, he became chairman of the Commission de l’Index, a commission set up in 1935 in order to reassure the critics of the way in which the index of retail prices was calculated43. On the occasion of his retirement, the catholic minister of industry and labour, Hendrik Heyman, characterised him as a “savant statisticien dont les chiffres, comme les procédés scientifiques dont ils émanent, pouvaient impunément affronter la critique”44.

23Armand Julin’s renown was not confined to Belgium. He was elected a member of the Institut International de Statistique in 1895, becoming its vice-president in 1929 and its president in 1936. Between 1921 and 1931, he was the Belgian delegate to the conferences of the International Labour Office. Julin was twice awarded the Heuschling price for statistics and he was an honorary member of numerous statistical societies in different countries45.

24Armand Julin can be regarded as a technocrat, who placed his expertise in techniques of social investigation at the disposal of a state apparatus whose cautious interventions seemed to provide the best guarantee for harmonious class relations. He was not only the father of the index of retail prices, but also one of its most authoritative and passionate defenders, which is not surprising given the strong symbolic value in Belgian industrial relations of a credible index figure46. Harmonious social relations were probably not Julin’s only motive for defending the index of retail prices. Since the index was conceived and developed within “his” administration of labour statistics, defending the index was tantamount to standing up for the prestige and reliability of the administration. It seems that both these considerations were present in Julin’s presentations of the budget inquiry of 1928-1929. Not only did he strive to document the advantageous social effects of Belgian capitalism, but he also emphasised Belgium’s excellence in the field of social statistics.

25The detached tone of scientific reporting that characterised Julin’s extensive article on the inquiry of 1928-1929 published in the journal of the Institut International de Statistique was absent from his other contributions on the subject, which focused on the standard of living of the Belgian worker and not on the budget inquiry as such. Julin drew comparisons with 1891 and the 1840s. Colourful details of investigations conducted in the 1840s were cited, illustrating the terrible conditions in which the Belgian proletariat had lived47. The message was clear: even those who were suffering from the present economic crisis could hardly imagine how bad things had been back in the 1840s.

26Parts of the latter article were reproduced and provided with additional material (mostly on the 19th century) in a discussion of the condition of the Belgian working class between 1830 and 193048. Again progress was emphasised, but the tone was less self-assured. This was understandable, perhaps, since the article was published at the end of the year in which the Belgian government had been forced to devalue the Belgian franc by 28%. Given the importance attached to a strong currency, this was regarded as a clear defeat after five years of deflationary policies. The quality of Belgian social statistics, however, remained uncontested, although Julin did formulate some minor critical remarks on 19th century (budget) inquiries. One has the impression, though, that the latter were merely mentioned in order to stress how impeccable the organisation of the inquiries of 1921 and 1928-1929 had been49.

28Armand Julin’s presentation of the results of the 1928-1929 inquiry in the Bulletin de l’Institut International de Statistique opens by situating this most recent exercise in the long and glorious tradition of Belgian (budget) inquiries starting in 1843. The quantitative richness of these inquiries is stressed. A comparison is made with Germany (1927), a far larger country, but where a proportionally smaller number of cases were collated than in Belgium50. On a second occasion, anticipating criticism of the relatively small number of white-collar workers involved in the inquiry of 1928-1929, Armand Julin repeated the comparison with Germany and extended it with a reference to the United States, again a large country with a proportionally smaller number of cases than Belgium. “Sans conteste, la proportion est bien plus satisfaisante en Belgique”, Armand Julin adds with satisfaction51.

29After assuring his readers that the sample is large enough, Armand Julin affirms explicitly that the inquiry was conducted according to “… ‘la méthode représentative’ dans laquelle on peut se contenter d’un nombre restreint d’échantillons pour représenter assez exactement le total”52. To us, such arguments seem self-evident. At the time, however, representativeness and sample taking were relatively new concepts that had only gradually been accepted since the turn of the century53.

32For the Leplaysians who developed and undertook the inquiry of 1891, the part measured had to consist of a collection of “typical cases”. In this respect, they were the inheritors of A. Quetelet’s holistic worldview, since “typical cases” formed the contingent manifestations of the one divine whole54. For every region and industrial sector, the inquiry of 1891 investigated a typical household: types moyens who embodied “la condition la plus fréquente”. In practice, families were selected in the light of their resemblance to the average cases that the organisers had in mind55. They showed a preference for large families whose breadwinners earned higher than average wages, lived in the most industrialised parts of the country and were often occupied in the best paid sectors56.

33The concept of representativeness as such was introduced a few years later by Anders Kiaer, a Norwegian statistician, who presented a paper on representative sampling to the Bern meeting of the Institut International de Statistique in 1895. As opposed to the Leplaysians, Kiaer did not aim to describe “typical cases” but a “variety of cases”. He wanted to give a fair view, not of a putative average, but of la variété des cas57. The newly named representative method met with scathing criticism from the adherents of full coverage: no calculations when observations can be made, was their motto58.

34At the Berlin session in 1903, a resolution was adopted acknowledging that a correct application of the representative method allowed generalisations. The report delivered by the Danish representative Adolph Jensen at the Rome session in 1925, revealed, however, that the representative method was still far from self-evident and that statisticians were still struggling with the question of how to compose a representative sample59. Between 1925 and 1934, the debate focused on the choice between either purposive sampling (“le choix judicieux”) or random sampling60. In 1934, the debate was concluded by the victory of the latter. The paper delivered by the Polish statistician Jerzy Neyman looked at Gini’s unfortunate experience with purposive sampling and turned most statisticians away from this sampling method. The combination of random sampling and stratified samples of the population became the orthodox method61.

35At first sight the former Leplaysian Armand Julin appears to have assimilated these latest developments. He refers explicitly to notions of representativeness and random choice (see further), two qualities that the inquiry of 1928-1929 was supposed to have. On the other hand, he stresses “la continuité de pensée et de méthode indispensable à la comparaison scientifique” between the inquiries of 1891, 1921 and 1928-192962. A belief in the continuity between the Leplaysian exercise of 1891 and the “modern” examples of la méthode représentative of 1921 and 1928-1929 was a precondition for Armand Julin‘s comparisons over time.

37The organisers of the inquiry of 1928-1929 wanted 2,500 families to contribute to the inquiry into food consumption, while 400 of these would also provide information on other expenses. From the experience with the 1921 inquiry, the organisers knew that quite a few participants would be lost in the course of the operation: only 56.5% of the 1,500 families involved in the investigation of 1921 had returned the complete information63.

38Although Julin also drew attention to the problem of non-response, he never mentioned how many families were “lost” in the course of the inquiry. The typewritten reports were quite explicit about it64. For the inquiry into expenditure on food which was held for two weeks in each season of the year, starting in May 1928 and ending in April 1929, the first period of observation ended with 54% of the forms returned. The second period of observation concluded with 47% completed forms, the third one with 43%, and the fourth one with 30%. After visits to the defecting families, some additional forms were returned: eventually 41% of the forms were completed. The inquiry into non-food expenditure, running from April 1928 to April 1929, required fewer participants but they had to co-operate for a whole year: 173 out of 400 families (43.2%) completed the exercise65.

39In order to create some enthusiasm for the whole operation, the Ministry of Labour had provided some incentives66. The agents enquêteurs were paid 20 francs for every form returned for the inquiry into spending on food and 200 francs for the whole of the inquiry into other expenditure. The families that co-operated were attracted by the promise that a sum of 50,000 Belgian francs would be distributed by lottery among the 100 families who completed the exercise.

40As the figures for non-response show, the possibility of financial reward does not seem to have convinced all the families concerned that it was worth the fuss and the intrusion in their private affairs. To the organisers of the inquiry, this did not come as a surprise. In their report they stated that the Belgian experience of 1928-1929 simply confirmed what was already known in Belgium and abroad: it was very difficult to convince families to participate and to continue to participate in such investigations67.

41Investigating families’ income and expenditure was a delicate matter, as the organisers of the inquiry acknowledged. Besides the promise of a financial reward, another way of overcoming families’ reluctance was by relying on people who would be trusted by the families. This procedure was not new. In the budget inquiry of 1891 for instance, the members of the Conseils de l’Industrie et du Travail were required to obtain the information from a working-class family known to them personally68. The inquiry by Louis Varlez into the incomes of Ghent working-class families around 1900 was carried out with the help of the local stewards of the trade unions69. The inquiry by the British Board of Trade in 1908 into the budgets of Belgian working-class families was also conducted with the unions’ help70.

42For the 1921 inquiry, “des enquêteurs habitant la région et parfaitement au courant du genre de vie des catégories sociales à étudier”, were engaged71. It appears that most of these were schoolteachers72. Persuading families to participate and to fill in the forms truthfully was something that required tact, skill and moral authority, according to the organisers of the 1928-1929 inquiry73. For this, the inquiry relied on the co-operation of directors of social organisations, members of the boards of sickness benefit funds, co-operatives, unions, teachers, priests, visiting nurses, inspectors of labour legislation and the agents who collected retail prices for the calculation of the price index74. In his statistics textbook of 1928, Armand Julin mentioned only the delegates from the socialist, Christian and independent sickness benefit funds75.

43These agents enquêteurs had to convince families that the results of the investigation would be published in generalised terms and that individual situations would be unrecognisable. This did not always help, however. Armand Julin speaks of “une répugnance marquée à livrer, même sous l’anonymat, le secret de leur vie de famille”76. The investigators were also required to stress that the inquiry would not serve as a basis for new taxes or would not give rise to wage cuts77.

44The agents enquêteurs were given some autonomy in the organisation of their work. They were the ones who had to select families among those who were prepared to lend themselves to the exercise78. They were also allowed to decide when exactly the four seasonal inquiries into food expenditure had to take place. The most appropriate moment had to be chosen, taking into account the families’ mood, the families’ time budget and local customs. Where possible a “normal” two-week period was picked, e.g. a fortnight without festivities that might entail extraordinary expenditure occasioned by family reunions79.

45As opposed to the inquiries of the 19th century that focused on blue-collar workers, the inquiries of 1921 and 1928-1929 explicitly included petty-bourgeois families. Both in the typewritten reports and in Julin’s and Gottschalk’s contributions, this was regarded as self-evident. No comment was made on the fact that in the Belgian tradition of family budget inquiries, this had not always been the case. It appears, however, that Belgian social policy, hitherto obsessed with the threat from labour, had at last “discovered” the petty bourgeoisie, regarded as a model of stability and therefore not a matter for urgent preoccupation. Before the First World War, complaints about the crisis of the petty bourgeoisie and its increased electoral importance since the changes in the suffrage (1893), led to a slight increase of interest in their fate80. In the post-war period, the wage-earning petty bourgeoisie was affected by the question of the price index.

46The Dutch/French typewritten report stated that of the 2,500 families whose co-operation was sought, 75% had to belong to the working class and 25% had to be members of the petty bourgeoisie81. Lion’s report simply took note of the fact that the proportion was the same as in the inquiry of 1921, but no other explanation was given82. The short presentation of the inquiry of 1921 does not give any clarifications either83. As the petty-bourgeois households were slightly less zealous in returning their forms on food expenditure, the proportion of working-class families rose to 78,3% for this part of the inquiry, but it fell to 67% in the part of the investigation dealing with other spending84.

47Nowhere in the discussion of the inquiry did anyone bother to define a working-class family. The term was simply circumscribed with the words “ménages d’ouvriers de l’industrie”85. Petty-bourgeois families were “bourgeois à revenus modestes”86 or in Dutch “klein-burgers met bescheiden inkomen”87. According to the typewritten reports these included white-collar workers’ families with a maximum annual income of 24,000 francs and retailers and industrialists employing a maximum of 10 blue or white-collar workers88. Farmers were explicitly excluded, given the difficulties in calculating their expenditure on food. Neither the unpublished reports nor Armand Julin himself specified what the composition of the group of petty-bourgeois families was exactly. Julin and Lion simply stated that most of the petty-bourgeois budgets came from white-collar workers89.

48The family was the inquiry’s unit of observation. It was defined as the assembly of persons united by family ties, living together and usually taking their meals together90. Families with lodgers were excluded. Collaborators were instructed not to bother about the families’ composition, given the fact that, at least where food consumption was concerned, results would be presented in units of consumption, the so-called quet91. Survey takers were invited to address mostly familles normales, defined as families with one or more children and without other dependants92.

49Families were also best chosen among “les familles originaires de la localité ou des environs, de façon à répondre au type moyen des ménages de la même catégorie professionnelle ou sociale”93. Here we recognise the Leplaysian ambition to gather information representing the condition of the alleged “average” worker. In his 1928 textbook on statistics, Julin was rather critical of this quest for la famille moyenne which he identified as a Leplaysian method that had nothing to do with statistics94. The inquiry carried out by his department in that same year, however, did exactly that: selecting “average” families.

51Where blue-collar workers were concerned, the inquiry had to reflect the relative importance of the number of workers employed by the different industries, according to the industrial census of 1910 and the industrial inquiry of 1926. The distribution of the number of workers per province in 1910 and 1926 was to dictate the geographical selection of families96. For the selection of petty-bourgeois families, the organisers admitted that they could only attempt to reflect the demographic strength of the Belgian provinces97. As early as the 1921 inquiry, an attempt at representativeness had been formulated. The brief details from this investigation do not reveal, though, whether any sectoral or geographical weighting was used in order to compose the sample98.

52How exactly the required distribution of cases was calculated is unclear. A comparison of the organisers’ “ideal” distribution with the results of the industrial inquiry of 1926 shows that they effectively translated the relative importance of the provinces as regards industrial employment and that they more or less succeeded in taking the sectoral distribution into account. Overall, one has the impression that the survey takers of 1928-1929 were aiming at producing some sort of proportionate stratified sample, in order to assure that the different strata of the Belgian working-class population (by province and by industry) would be correctly presented99. Within these strata, however, the agents enquêteurs were led not by random sampling, but rather by a variant of Leplaysian selection. It seems that the inquiry of 1928-1929 showed a mixture of “old” and “new” sampling methods and that the survey takers themselves had anything but a clear view on the matter.

53In this respect, the international conferences of labour statisticians were not of much help. According to Armand Julin, the inquiry had carried out the recommendations of the third conference of labour statisticians (1926)100. Their selection advice, however, appears to have been very general:

55In their world-wide survey of family budget inquiries published almost ten years later, Williams and Zimmerman pointed out that the adherents of the statistical approach to family budgets had criticised the Leplaysians for choosing supposedly typical families, while they were using unsystematic selection methods themselves103. In his methodological discussion of family budget inquiries, held in some thirty countries in the 1920s and 1930s, Morse Woodbury declared that, although the question was fundamental to claims of representativeness, most inquiries gave little or no information on their selection procedures104.

57The industrialised provinces of Hainault and Liège, which were supposed to cover 44% of the families in the part of the inquiry dealing with food consumption, in fact included 59.5% of the participants. Within the province of Liège, the single Liège arrondissement, containing the city of Liège itself, provided 79% of the province’s contribution. This means that one single Belgian arrondissement (out of 41) provided 23.6% of the households participating in the “national” budget inquiry. One may add that the organisers’ “ideal” distribution was itself already somewhat biased. It was based on the industrial inquiry of 1926 which covered only firms employing ten workers and more. This led to an overrepresentation of firms in the provinces of Liège and Hainault, whose size was generally bigger106.

60The organisers of the budget inquiry of 1928-1929 intended to gather a sample that would reflect sectoral and regional differences within the Belgian working class. The use of census data as a basis for the composition of their sample illustrates that they were familiar with the most recent methods in sample taking. The tone of the comments on this topic published in the first half of the 1930s reveals that this approach was still far from self-evident109.

62The unpublished reports fully acknowledged that the proposed distribution of cases had not been reached at all. Immediately following this confession, however, it was argued that there was no reason to split hairs. Downplaying the original goal, the report suggested that, apart from the nombres fictifs ayant servi de base au travail, the inquiry might be considered representative of the way of life of the Belgian working class111. That the organisers of the inquiry should first attempt to produce a stratified sample and then, after failing to do so, should finally decide that this was not so important after all, is again symptomatic of the pragmatic attitude of labour statisticians in this respect.

63In his account of the inquiry, Armand Julin did not mention that the organisers had been confronted, in their own terms, with a problem of representativeness. Although he acknowledged the problem of non-response and even complained about it, he never alluded to the fact that this might have influenced the composition of the sample. In fact, Julin remained rather discrete as to the methodological issues of the whole inquiry. He did not mention the strong geographical bias of the investigation. Bending the truth as to what had been achieved, Julin asserted that the branches of industry were represented in the inquiry selon leur importance112.

64More remarkable still is Julin’s attempt to see as an advantage what had in fact been considered a serious setback by the organisers, viz. the fact that results did not match the “ideal” blueprint. Julin’s magic word in this attempt was le hasard (chance). This referred to the new method of sample taking, i.e. random sampling, and the suggestion was that the Belgian statisticians had in fact applied la méthode représentative in which le hasard played an important part113. On three successive occasions, Julin argued, “chance” (i.e. a random selection) had taken place. First of all, survey takers had complete liberty as to what blue or white-collar worker’s family they would choose, in the second place a lot of families had failed to complete the survey and, finally, the administration eliminated those families who had not provided adequate information114!

65Armand Julin did not seem to be aware of the fact that the principle of random selection had not been respected because selection was done by human choice and “representative” cases had been chosen. In his 1928 textbook on statistics, however, Armand Julin appeared to be rather critical of any intervention that might arbitrarily influence the composition of the sample115. Moreover, the element of random selection, if any, was bound to be undermined as a result of the fact that better-off families had more chance of being selected to participate in such exercises. They were more likely to be members of the same organisations as the survey takers and they were less likely to be eliminated for not filling in the forms properly. Statisticians of the 1920s were already aware of the fact that housewives who undertake to keep an account will as a rule have a greater sense and understanding of economics than the general run of housewives; and these individual qualities will influence more or less the manner in which the family distributes its expenditures116. In contrast to his light-hearted comments on the problem of non-response with regard to the 1928-1929 budget inquiry, Julin acknowledged that lack of responses undermined the whole idea of random sampling in a methodological article written in 1930 and published two years later117.

66In the 1891 budget inquiry, combining various criteria led to the selection of large families: the average number of persons per household was 6.2 while the Belgian average was 4.5118. This bias towards large families was less pronounced in the 1928/29 survey. Here we find 4.4 persons per family in the blue-collar households and 3.7 persons per family in petty-bourgeois families, while the Belgian average was 3.4. Apart from the proposal to select “normal” families with children and no other dependants, the organisers did not provide any explicit instructions. Although the nuclear family remained predominant (78.0% of cases in blue-collar workers’ families and 65.2% in petty-bourgeois families), we find a significant number of couples without children and “normal” couples with children but with other dependants. The fact that organisers and their co-operators had so many difficulties in convincing people to continue to participate in the investigation might have made them less demanding.

67As opposed to his discretion with regard to the geographical and sectoral composition of the sample, Julin explicitly addressed some of its demographic features, such as the numbers of children in the selected families. In order to assure the reader of the representativeness of the sample, he compared the number of children per family with the statistics with regard to families receiving family allowances: the two appeared to be a perfect match119. Comparison with the results of the population census of 1920 confirms that the inquiry adequately reflected the distribution of families with children, but that childless families were underrepresented. In 1920, 27.3% of married couples had no living children, while married couples participating in the inquiry appeared to be childless in only 13.4% of cases120.

69Families participating in the inquiry into food expenditure carried out for two weeks in each of the four seasons were given a 14-pages carnet de ménage for each period of observation, in which they were supposed to write down daily what their expenses were. The cover of the booklet was reserved for information regarding the composition of the family, occupations of its members, address and the family’s income sources. Those who also participated in the inquiry into other expenses, received a simple carnet de notes for each season of the year. Day after day, the page on the right was to register income, the page on the left expenditure. To the organisers, it was obvious that the housewife would be the one keeping the books121.

70The agents enquêteurs had to make sure that the entries were exact and were filled in on a regular basis. They had to check carefully whether the earnings of all family members were registered, and that additional resources and income in kind were not forgotten122. In order to keep the agents enquêteurs alert they were given a list of income sources a family might possibly have123.

71During the processing of the data, doubts arose with regard to the accuracy of some of the income figures. The agents enquêteurs were asked to check the information. The housewives concerned were invited to explain the anomalies. The seasonal differences in income could in some cases be explained by wage increases. Unemployment, local festivities, strikes, bad weather, overtime and sickness were also responsible for some of the divergences124.

72In some cases co-operating housewives had overlooked certain supplementary resources, such as family allowances, pensions, insurance, income earned in a small business, a second job or overtime, free coal and free lodgings125. These omissions were not merely accidental. It appears that families had few problems revealing what they ate, but providing information on their overall resources was a more delicate matter.

73In general, the Dutch/French report concluded, income from wage labour was reported accurately127. When wages were unspecified or doubtful, survey takers compared them with wages registered for another season of the year or wages earned by members of the same occupation128. On the basis of such supplementary information, some of the budgets were “corrected”129.

74When information on such items as family allowances, pensions or free lodgings was recorded in one of the four periods of observation, but missing in the others, the information was added if no changes in the situation of the family had occurred130. The same procedure was followed when in the supplementary inquiry into other expenditures, an investigation lasting for a whole year, these sources of income had been omitted. Again assuming that it was safe to extrapolate, the average of the four seasonal figures was multiplied by 26131. Sometimes the families had completely forgotten to mention any secondary sources of income. As a result, the agents enquêteurs were obliged to formulate numerous requests for additional information132. It remains unclear whether these were successful.

76In his published report, Armand Julin commented on the difficulties encountered in collecting information about resources, but he did not mention adjustments or extrapolations. He suggested that the compilers of the data had been quite strict and had preferred to collect a limited number of observations, providing they presented “toutes les garanties d’exactitude et de sincérité”133.

77When budgets revealed a deficit, survey takers felt more encouraged to persist in their quest for additional information on household resources: families had to explain how they could possibly cope134. Quite remarkably, the inadequacy of information about secondary and irregular sources of income was not considered a possible explanation for the “deficit”. Survey takers assumed that other supplementary resources were available and savings or shares seemed the most likely ones. A major problem, however, was that the participating families never mentioned such resources135. Questioned on the subject, they always succeeded in avoiding the issue and sometimes they simply refused to discuss the matter any further. Even Armand Julin, who tended to minimise the inquiry’s shortcomings, admitted that the investigation offered little information in this respect, which he attributed to the fact that:

78Fear of the tax administration was mentioned as one of the reasons that withheld them from disclosing this information137.

79Both the typewritten reports and Julin suggested that savings and the selling of shares could have covered the deficits138. It is not surprising that they should suggest that families had savings: savings banks were booming in the inter-war period139. More remarkable is the assumption that ordinary people would have shares. Unfortunately, the idea was not developed any further.

81We must therefore be aware of the fact that the 1928-1929 family budget inquiry largely represents the situation of the working class in the provinces whose industrial workers were the most prosperous141. Another indication that the industrial workers’ families who took part in the survey were relatively well off is that 15% of them had a yearly income that was equal to, or surpassed, the 24,000 francs maximum limit for the selection of petty-bourgeois families. 30% of industrial workers’ families earned as much as, or more than, the average white-collar worker’s family included in the inquiry142. In his international survey of budget inquiries, a collaborator of the International Labour Office admitted that budget inquiries in general only touched the most prosperous working-class families:

87Where Belgian budget inquiries are concerned, the composition of family income and the contributions of its different members were explicitly addressed only in the 1890’s. This is not a coincidence, probably, since the uprising of 1886 encouraged commentators to emphasise, even more than before, the connection between girls’ and women’s paid labour and the social crisis that Belgian society was going through145. Armand Julin tackled the question of the composition of family resources in his discussion of the budgets of 1886 and 1891. He dealt with this in relation to the increasing standard of living, the alleged disintegration of the family and the welcome marginalisation of women’s work146. A more nuanced view is presented by Louis Varlez, who did not in fact conduct a budget inquiry but focused entirely on the level and the composition of family resources147. Although he also cherished the idea of a single, male breadwinner, he acknowledged that in the textile city of Ghent, characterised by a low male wage level and a lot of job opportunities for women and children, all family members made an important contribution to household income.

88By the 1920s, the question of women’s work in relation to the family budget seemed to have become marginal. In the discussion of the budget inquiries of the 1920s, most attention was paid to expenditure in general and food expenditure in particular. This is not surprising, since these investigations had been explicitly designed to collect information on patterns of consumption. Armand Julin, who in the 1890s had an axe to grind concerning married women’s paid labour and especially factory work, ignored the matter completely in his discussion of the 1928-1929 inquiry. In contrast with the 1890s, it did not seem to be necessary to convince the reader that real wages had improved to such an extent that women’s work had become “superfluous”. Instead Armand Julin focused on the relationship between blue-collar workers’ income levels and patterns of expenditure and the fact that the differences in the ways of life of industrial workers and white-collar workers could not be explained by income variations alone148.
   

89Social investigations have never been disinterested activities, but with the amplification of social policies after the First World War, they became an integrated part of the arsenal of policy instruments. Together with their political relevance, the qualitative demands made on statistical inquiries rose. More than before, it became crucial to be able to demonstrate that reality had been mapped in a fair and impartial way. The aim developed of being representative, objective, even scientific149. Survey takers tried to live up to these expectations, but work in the field hardly matched the theoretical requirements of scientific worth and representativeness. Technocrats, such as Armand Julin, who were as much devoted to “their” social statistics as to the social policies that they served, found it difficult to assess the shortcomings of social statistics. Openly questioning the figures could have engendered unwanted side effects for the legitimacy of the policies involved. As a result, the difficulties encountered in investigations like the 1928-1929 budget inquiry were dealt with more than leniently: they were downplayed or concealed. Biased as they were, they were nonetheless presented as representative surveys of the wage-earning population.