1The censuses of 1851 and 1872 included a question on religion which was omitted in 1856 and subsequently removed. Claude Dargent examines the “quite severe difficulties” mentioned in the 1851 census report. Comparing the estimated numbers of Protestants in 1851 and 1861, and drawing upon the results published for cities such as Paris and Bordeaux, he shows that the 1851 statistics are very probably inaccurate and postulates a reason for this. Under the Concordat, the payment of priests, pastors and rabbis was dependent upon a minimum congregation size. In 1851, the figures available to the government administration may have been severely under-estimated – since many Protestants did not report their religion in the census – and inaccurate results may have been published to protect clerical stipends. This research paves the way for a more specific historical study of the “vicissitudes” of census data publication, that could be usefully associated with an analysis of archive material from the relevant administrations.

2In response to the grave misuse of public statistics from the 1930s to identify people of Jewish origin, several countries, including the United States, stopped asking questions about religion in censuses after the Second World War. In France, this decision was taken in the nineteenth century, in the early days of the Third Republic. It ended a brief period of just twenty years or so, the general population census of 1851 having been the first to include a question on religious affiliation. At that time, out of a total population of 35,783,000, France counted 34,931,000 reported Catholics, 775,000 Protestants, 74,000 Jews and 3,000 persons with no religious affiliation (Statistique générale de la France, 1855, p. xxv). The question was asked again in 1861, 1866 and 1872. From then on, questions on religious affiliation were only ever asked in the Haut-Rhin, Bas-Rhin and Moselle départements between 1919 (when they were returned to France) and 1965. In these départements, the separation of Church and State enacted in 1905 did not apply, so the Concordat remained in force.

3Some scholars interested in French religious practice mourn the disappearance of this information source. Clearly, these data were no longer of much interest to the public authorities after the Separation Act of 1905. Since the state no longer paid any clerical stipends, there was no further need to count congregations in order to apportion budgets among the various religious denominations. In this regard, one could claim that it is no more justified to ask residents in France about their religion than about their membership of clubs or associations or their political leanings – data which are useful for describing French society but which are by no means vital to the government. Entering a universe perceived in France as pertaining to individual conscience may affect the respondents’ propensity to reply sincerely to other questions in the census, thereby threatening its reliability. Such a consequence is clearly undesirable.

4Contrary to appearances, and assuming indeed that it is founded, [1] the notion of a “return to religiosity” that has developed in recent years, is irrelevant here. In Europe, the religious revival often manifests not in a return to the church – either traditional or modern – but in a resurgence of beliefs (Lambert, 2004) whose “natural decline” was previously seen as irreversible. This is consistent with the trend already identified by Grace Davie, and summarized in the formula believing without belonging (Davie, 1994). To capture this ongoing movement, research must focus on representations, a task for which the census is clearly not the most appropriate instrument.

5Classic survey techniques based on representative samples are much more suited to the study of religiosity, whatever their well-known imperfections. For minority religious beliefs, however, this approach raises specific problems: secondary analysis of ordinary national samples is impossible for reasons of sample size and hence of uncertain data reliability. The only way around this difficulty is to resort to aggregated surveys [2] and/or very large national samples, [3] since the high cost of identifying respondents belonging to a rare religion from among the general population makes it difficult to constitute specific samples. Moreover, in the absence of periodic censuses, we have no information on the social structures of the populations concerned, yet these social structures form the basis of the sampling quotas habitually used in France.

6These real difficulties doubtless explain why the disappearance of the “religion column” in the French census is so sadly lamented, notably by specialists of minority religions, for whom surveys are now the only available tool. Taking an opposite perspective, which now becomes understandable, Jean Baubérot qualifies the four nineteenth-century censuses that collected religious data in France as “life-size surveys” (Baubérot, 1988, pp. 21-24). Under this denomination, these censuses are seen as a sort of random sample survey free from the element of uncertainty described by the theory of probabilities. Likewise, Philippe Chareyre considers that the 1851 census “can serve as a fairly reliable source of official statistics” (Chareyre, 2002, p. 148). [4]

7This nostalgia for the figures once provided by public statistics reflects an all too rare concern to base social analysis on reliable foundations and, as such, deserves to be acknowledged. Yet it carries a risk of giving too much credence to these nineteenth-century census data. In fact, as we shall see, a closer look at these enumerations reveals that the indications they provide in terms of religious affiliation, concerning Protestants especially, must be treated with great caution. At the same time, however, by re-examining these data, we shed new light on the complex relations between statistics and state power, two concepts whose links extend far beyond etymology alone (Desrosières, 1993).

I – The 1851 census and religions: a case of self-censorship

8In 1851, 2.16% of French inhabitants reported being Protestant. They included 481,000 members of the Reformed Church, 268,000 Lutherans and 26,000 “others”, mainly members of small Protestant denominations (Statistique générale de la France, 1855, p. xxv). In the census of 1861, they totalled 802,000, with 480,000 Reformed Church members, 282,000 Lutherans and 40,000 members of other Protestant denominations (Statistique de la France, 1864, p. xiv). In 1866, out of a total of 847,000 Protestants, 516,000 were Reformed, 287,000 Lutheran and 44,000 other denominations. [5] Last, in 1872, the loss of Alsace-Lorraine brought the number of Protestants down to 581,000, with 468,000 Reformed 80,000 Lutheran and 33,000 “others”.

9To begin with, we must ask what these figures measure exactly. The question asked was: “What is your religion?”. While the summary tables classify the collected data under the heading “religious statistics”, the census question makes no explicit reference to membership of a particular church. From the viewpoint of theology and Protestant ecclesiology, this is a good thing: if one cannot be a good Catholic without going to church, since attendance at mass is a canonical obligation, the same rule does not apply to Protestants. When, in line with the sociology of Catholicism practiced by Gabriel Le Bras and Fernand Boulard, a sociology of Protestantism was first evoked, this was one of the first questions raised. Fervently upholding the identity of their church, Protestant academics point out that it cannot be correctly described on the basis of the Catholic definition (Léonard, 1959, p. 130; Mehl, 1962, pp. 10-11). [6] I have shown elsewhere (Dargent, 2005) that this specifically French reticence owed much in our country to the memory of the Church of the Desert (following the revocation of the Edict of Nantes, when Protestants were condemned to isolation, with the exception of episodic services held in certain bastions of the Reformed Church). The question asked in the nineteenth-century censuses lifts this objection as it does not prevent detached believers who do not attend church but who read the Bible and act in accordance with the Gospel precepts from reporting themselves as Protestant Christians. Although this precaution is essentially a academic issue today, the question used in the census is able to satisfy the most demanding followers of this tradition.

10The selected approach being pertinent even with respect to Protestant theology and ecclesiology, we can proceed to a fuller examination of the reliability of collected data. The census instrument itself is not brought into question here. Admittedly, during the first half of the nineteenth century, its first applications were tainted with serious bias (van de Walle, 1974) that invite precaution in the analysis of census data from this period (Bonneuil, 1997, p. 6). But from 1851, the French administration introduced improvements, rendering the technical quality of reported data more acceptable from the mid nineteenth century. [7] The overall population figures obtained from the censuses are plausible, even if the mayors responsible for the operation introduced inaccuracy by spreading data collection over one or two months.

11But can the same be said for the breakdown of the general population by religious affiliation? We know that the capacity of the nineteenth-century censuses to classify French people by occupational categories – a much more complex operation than simple enumeration – has been criticized (Marchand and Thélot, 1991, pp. 8-11). This concern does not appear relevant to our study, however, since religious categories are in principle quite distinct and do not face the same problems of overlapping boundaries and names encountered for occupations and economic sectors.

12Although É.-G. Léonard suspects the existence of “‘omissions’ on the part of clerical employees and mayors” (Léonard, 1953, p. 22), this problem is very likely to be marginal. Mayors were not required to estimate the religious affiliation of inhabitants, but to ask them about this subject, like all the others. Given the pressure exerted by the central authorities to ensure the accuracy of reported figures, it is therefore difficult to believe that the data transmitted to Statistique générale de la France (SGF) were substantially falsified.

13The quality of the information collected deserves close scrutiny, however. In particular, much closer attention must be paid to an anomaly too rarely studied up to now: after their innovative introduction in 1851, questions about religious affiliation disappeared from the census of 1856. According to some, they were removed in response to protests over the inclusion of a such a question category (Dupâquier and Le Mée, 1988, p. 39). But this explanation does not account for the reintroduction of a question on religion in 1861. We note, on the other hand, that the religious data collected in 1851 did prove highly problematic for the enumerators (Poulat, 1956, p. 24). I would thus like to explore a different hypothesis: might these very difficulties not have prompted the removal of questions on religious matters in 1856?

14Curiously, indeed, only the overall results collected in 1851 were published, along with those of the prefectures and sub-prefectures (Statistique générale de la France, 1855, pp. 263 to 279). This choice is justified in a way which raises more questions than it answers: “due to considerations of a particular nature, the administration has decided not to print the table giving the distribution of religious denominations by département” (Statistique générale de la France, 1855, p. xxv). The author gives no further information on the nature of these “particular considerations”. Yet, the withholding of data in this manner, accompanied by such an elliptic explanatory note, is no ordinary occurrence in this type of publication.

15As the persons in charge at the time left no subsequent explanations, we would be faced with a still unresolved mystery had the publication devoted to the 1861 census not returned to the question some years later. The change of tone is radical. We read that: “This is the second time that religious denominations have been included in the French census, and we are pleased to say that the quite severe difficulties encountered by this special enumeration process in 1851 did not recur in 1861. Nonetheless, as the results collected over this last year still cannot be viewed as a faithful expression of the truth, they should be accepted with a certain reserve” (Statistique générale de la France, 1864, p. xiv). And five years later, in 1866, the author once again expressed satisfaction that the difficulties of 1851 had not recurred.

16The administration never expanded upon these cryptic remarks, and the 1872 census was the last to include questions on religious affiliation. The published statistics on religion must therefore be handled with prudence, as explicitly recommended by the authors. But I would like to dig a bit deeper here: what might be the reasons for this self-confessed administrative censorship?

17In 1851, it was above all the local distribution of religious denominations which caused problems – hence the non-publication of population statistics by religious affiliation at département level. Second, when an administration mentions, with no further explanation, the need for secrecy due to “considerations of a particular nature”, one can reasonably assume that the problem is political, especially when the obstacle encountered in 1851 is retrospectively – and laconically – described ten years later as posing “quite severe difficulties”.

18We should bear in mind the relationship between religious denominations and political power in the nineteenth century. Under the Concordatory system established in 1801-1802 with the Catholic, Lutheran and Reformed Churches, and some years later with the Jewish synagogues, the salaries of priests, pastors and rabbis were paid by the state. For some sixty years, therefore, the clergies of these different religions had been remunerated on the government budget, providing that their congregations totalled 6,000 persons or more. Perhaps it is here, at the crossroads between demography, religion and politics, that the answer to our problem can be found.

19First, it would seem that the Jews are not concerned. In mid-nineteenth century France, many of them were city dwellers, and we have seen that the results obtained for the urban population were published for the prefectures and sub-prefectures, so did not pose any problems. If we exclude the small numbers of Jews who spread across Provence from the Comtat Venaissin, the only French rural Jews at that time lived in Alsace, where they were scattered among 187 municipalities. It is difficult to understand, therefore, how the results of the two départements along the Rhine could be considered sufficient to explain a publication decision which concerns the whole of France.

20This leaves the Catholics and the Protestants. We know that the minimum congregation size of 6,000 was set in relation to the Catholic Church, whose parishes counted 8,500 to 9,000 Catholics on average. A threshold of 6,000 thus ensured that most parishes had their own priest. So there is no reason for the Catholic Church to be the root of the problem, since it served as the reference, with 97.6% of French people reporting themselves as Catholics in 1851.

21This minimum congregation size of 6,000 is highly artificial for the Protestants, on the other hand (Robert, 1961, pp. 71 seq.). It is nonetheless the prerequisite for forming a consistorial church and receiving state remuneration for the pastor. According to figures given at the time by the Reformed Church, around eighty pastors had been receiving government stipends for the past half-century.

22This suggests the following hypothesis: the public authorities failed to publish the local breakdown of religious denominations because they saw that the small numbers of Protestants counted in many départements did not justify the existence of the consistorial Churches established there. The public disclosure of this state of affairs would have sparked a whole series of politico-religious problems, starting with accusations of favouritism towards the Protestants, notably on the part of the Catholics, with whom Napoleon III was on uncomfortable terms, as we shall see. Though we tend to forget the degree of antagonism between the two Christian churches in France at that time, this would have been quite enough to rekindle the conflict between them.

II – Protestant populations from 1851 to 1861: growth in mission areas and stagnation in bastions. A real or fictitious trend?

23In support of the above hypothesis, we note that local monographs have sometimes signalled a disagreement between the census figures concerning the prefectures and sub-prefectures (the only published statistics) and data from other sources. In Bordeaux, for example, the 1851 census enumerates 2,095 Protestants (2,008 Reformed and 87 Lutheran), while a survey by the consistory in 1857 counts 3,400. And when consulted on the question, the consistory invariably gave a figure of 6,000 to 8,000 Reformed Church members, alleging that the Protestants who attended church did not come exclusively from the city of Bordeaux, but also from the surrounding countryside (Pacteau de Luze, 1976, pp. 201-202). Congregation size was thus large enough to justify state funding of a pastor’s stipend in Bordeaux. However, given the figures of the 1851 census, this presupposes that 4,000 to 6,000 Reformed Church members came regularly from the rural cantons to worship, a considerable and suspiciously high number given the means of transport available at that time.

24This leaves us with two options: either the Protestant population estimates serving to legitimize the existence of this consistorial Church were exaggerated, or the 1851 census severely underestimated the number of Reformed Church members in Bordeaux because many did not report themselves as such. The subsequent figures appear to validate the second hypothesis. The 1861 census enumerates 6,336 Protestants (Table 1), [8] indicating a three-fold increase in their population in just ten years! The city’s population did increase between the two dates, but to a much smaller extent, rising from 130,927 to 162,750 inhabitants (+25%). From then on, the Protestant and general populations grew at very similar rates, and most of the gap attributable to the 1851 under-estimation was absorbed.

Table 1

Religious affiliation in Bordeaux according to the censuses of 1851 to 1872

Year Population Growth since previous census (%) Catholics Protestants Jews Other religion, not reported or no religion Total population Protestants Total population 1851 126,399 2,095 2,319 114 130,927 1861 153,722 6,336 2,692 – 162,750 +202 +25 1866 184,381 7,444 2,416 – 194,241 +17 +19 1872* 181,679 8,558 2,716 1,802 194,055 +15 +0 Source: Censuses, Statistique générale de la France. * The sum of members of different religions should give a total population of 194,755 inhabitants. The transcription error, albeit minor, is probably in the number of Jews, which should perhaps read 2,016, or in “other religion, or religion not reported” which should perhaps read 1,102.

Religious affiliation in Bordeaux according to the censuses of 1851 to 1872

25A similar pattern is observed in many French cities. In Rouen, 538 Refor-med Church members and 150 Lutherans were enumerated in 1851, then 1,114 Protestants in 1861, for a population which grew by just 2%. In Marseilles, 3,252 Protestants were counted in 1851, and 9,126 ten years later (+180%) while the city’s population grew by less than one-third. In Toulouse, the 1951 census counted 490 Reformed and 44 Lutherans. In 1861, 1,574 citizens of Toulouse reported themselves as Protestants in a city that had grown by just 20%.

26One might argue that these figures reflect rates of rural exodus that differed by religious denomination. [9] Yet this explanation is clearly inadequate to account for the observed growth in the urban Reformed population – unless we make the implausible assumption that the religious variable is a major explanatory factor of geographical mobility. And in any case, this interpretation cannot account for the larger contrast observed in smaller towns. Wherever Protestants form a minority, we observe a recurrence of the process described above for cities. In Cahors, the number of reported Protestants rose from 14 to 46 between 1851 and 1861 while the population grew by just 4%. In Saumur, the census counted 116 Reformed Church members – and 1 Lutheran – in the mid-nineteenth century, rising to 173 ten years later, while the overall population remained unchanged. Even in Niort, a town in the historically Protestant region of Poitou long prey to severe religious persecution, the apparent increase is considerable, with Protestant numbers rising from 384 to 2,400 among a population that grew from 18,727 to just 20,831!

27By contrast, in places where the Protestant community was long-established, there is little difference between the two census figures, and sometimes even a decrease, probably linked to migration. In Die, 1,117 Reformed Church members were counted in 1851, and 1,116 in 1861; in Orthez, the number of Protestants fell from 1,252 to 1,211 between the two dates, in Uzès, from 1,772 to 1,643 and even in La Rochelle, from 864 to 700. In these towns, where Protestants were a clearly identified and locally prominent historical community, there is no increase between the two dates, since there was no dissimulation in 1851. Everywhere else, the increase between the two censuses was huge and offered clear evidence of initial under-reporting.

28Hence, if the Reformed city-dwellers concealed their affiliation in 1851, except in the rare city bastions where they formed a majority or a large minority, there is good reason to believe that the same was true in the rural areas where the vast majority of the Protestant population lived at that time (Encrevé, 1985, pp. 34-35). If this were indeed the case, then it was impossible to publish results by département without revealing that across large parts of France the local Reformed populations were too small to justify the existence of a church and hence the stipend of a pastor.

29We must also consider the hypothesis whereby the 1851 census administrators were confronted by a substantial under-estimation of the Reformed population. Given its scale, this phenomenon cannot be attributed simply to omissions by clerical mayors, but is probably due to a reluctance on the part of Reformed Church members to report their affiliation. This is hardly surprising. Of course, the context was no longer that of the eighteenth century following the revocation of the edict of Nantes in 1685. But the Edict of Toleration of 1787, which restored Protestants to their civil rights and recognized their presence on French soil, was only sixty years old in 1851, and the White Terror [10] was very recent. The memory of persecution remained acute among the Protestants, so they may indeed have been unwilling to inform a state representative of their affiliation to a religion that had so recently been proscribed.

30The sums tend to support this hypothesis. Though no breakdown by département is available, we do have an overall figure of 481,000 for French Reformed Church membership. If this “official” 1851 total of 481,000 is divided by the 81 consistories existing at that time, we get 5,938, almost the exact minimum congregation size imposed by law. This is a somewhat troubling coincidence, and it prompts the following, complementary, hypothesis: confronted with significant under-reporting of Reformed Church membership in France, the authorities artificially increased the overall figure in order to justify the continued existence of the country’s 81 consistorial churches. But to ensure discretion, they did not publish local statistics, which would have made the adjustment rather too visible, given that data for rural municipalities were collected under the authority of the local mayors.

31Subsequent data confirm this hypothesis. In the light of the difficulties detailed above, we can easily understand why the public authorities, and hence the French official statistical service, removed all questions relating to religious affiliation from the 1856 census. So why did the question recur in 1861? In that year, the Ministry of Public Instruction and Religion wished to make a specific enumeration of religious communities. [11] This made it difficult to avoid asking the question about religious affiliation to the French population as a whole.

32The census revealed that between 1851 and 1861, the number of Reformed Church members had fallen slightly, from 481,000 to 480,000. This stagnation itself raises a number of questions. [12] Over this same period, the French population grew by 4.5%. Contrasting with this rapid demographic transition, the stationary Reformed population is all the more surprising given the subsequent reversal of the situation: from 1861 to 1866, the enumerated Reformed population grew four times faster than the population in general, at a rate of 7.5% versus just 1.8%!

33However, if the hypothesis advanced here is correct, this change becomes much more comprehensible. As the Reformed population figures published in 1851 are artificially high, those collected ten years later do not signify rapid growth but rather a reduction in under-reporting. The Second Empire was hardly a golden age for Protestant evangelization in France (Baubérot, 1988, p. 24), so the change between 1861 and 1866 is doubtless due not to a massive conversion campaign, but rather to a further reduction in under-reporting – occurring not so much in the towns, where the 1861 figures became, as we have seen, more plausible, as in the countryside, where Protestants had remained mistrustful, and in the Paris region, as we shall see.

III – Historical comparisons and the particularities of Paris

34The rapid increase in Protestant populations enumerated in the cities explains the satisfaction expressed by the Statistique générale de la France in 1861. The Reformed Church members were now less wary of revealing their affiliation to the census agent, thus making the département statistics more “presentable”. The logical explanation is that the French were becoming accustomed to the census: the troubled political times of the Second Republic were now well past, and the fear of religious persecution was waning.

35Beyond this general acceptance of the census, can it be said that the Second Empire, with its tolerant religious policy, reassured the French Protestants? This hypothesis is tenuous. The policy of Napoleon III in favour of Italian unification certainly compromised his relations with the Pope, but the Protestants, for their part, had many reasons of their own to be dissatisfied with the regime. The main bone of discontent was the renewed requirement for authorization to hold meetings, even religious ones. This raised numerous problems for the organization of religious services since, unlike Catholic priests, not all French pastors had access to a church building. Indeed, these difficulties may have encouraged Protestants to officially report their affiliation to a religion whose members were prevented from gathering freely in certain regions.

36The number of reported Protestants was large enough, in any case, to safeguard the 81 Reformed Churches in France, so département-level statistics could be published (Statistique générale de la France, 1864, pp. 84-87). Admittedly, the threshold of 6,000 churchgoers per consistory was not always reached, but such cases were exceptional [13] and the overall average was sufficient – although only just.

37It is also instructive to compare the years 1861 and 1866: the département data reveal that growth in the urban Protestant population cannot be explained by rural exodus alone. The départements with a substantial Protestant population were agricultural ones in which massive rural exodus led to a decline or, at best, stagnation of population between 1861 and 1866. This was the case in Tarn-et-Garonne (–1%), Drôme (–1.3%), Basses-Pyrénées (–1.2%), Seine-Inférieure (+0%) and Tarn (+0.3%). Yet, far from declining, the Reformed population increased in these départements, sometimes quite sharply (respectively +1%, +7.1%, +10.9%, +3.1% and +3.9%) – proof that the reduction in Protestant under-reporting was still having visible effects. [14] Since the rural Protestant population increased faster than that of Catholics, rural exodus cannot explain the increase in Protestant city-dwellers.

38This points to initial under-reporting of Protestants in 1851 – that was absorbed in subsequent censuses. [15] A similar phenomenon has been observed elsewhere by historical demographers. For example, in the aftermath of the 1789 revolution, it was the Catholic regions which refused to comply with the new civil registration procedures that replaced the parish registers kept by priests. Some decades later, the 1841 census sparked popular protests against tax reform in certain départements. In fact, although there were no plans to introduce a new tax, despite rumours to that effect, one aim of the census, instigated by the minister of finance Georges Humann, was indeed to boost the income raised from direct taxes (Ploux, 1999). And beyond national borders, Morgane Labbé described a similar, more explicitly political protest on the part of the Albanian minority in Kosovo with respect to the Serbian majority, which took the form of under-registration of births followed by a boycott of the 1991 census (Labbé, 2000, pp. 86 seq.).

39There are nonetheless two important differences between the Protestant under-reporting in 1851 and this more recent episode. First, the Protestant were not seeking to attract attention to themselves. Their aim was not to make a visible show of civil disobedience but, on the contrary, to remain as discreet as possible. Second, the Protestant under-reporting probably did not manifest as a refusal to take part in the census, which would have affected the overall population estimate for France, but more simply as a concealment of religious affiliation.

40We will now take a closer look at the case of Paris (Table 2). The propensity to report Protestant affiliation also increased in Paris in the 1850s, but less rapidly than in many other provincial cities. But unlike these other cities, the increase continued at the same pace from 1861 to 1866, with the Protestant population growing by 75% compared with just 8% for Parisians in general. The population breakdown by religious affiliation published for the Seine département, confirms this rapid pace of growth. The population of reported Reformed Church members more than doubled between the two dates, from 9,584 to 19,904 (an increase of 108%), while the population of the département grew by just 8%. Why did growth continue for longer than in the provincial cities?

Table 2

Religious affiliation in Paris and in the Seine département^(a) according to the censuses of 1851 to 1872

Paris Year Population Growth since previous census (%) Catholics Protestants Jews Other religion, not reported or no religion Total population Protestants Total population 1851 1,025,169 13,366 10,719 4,008 1,053,262 1861 1,656,661 23,585 14,867 1,028 1,696,141 +76 +61 1866 1,757,823 41,244 20,615 5,592 1,825,274 +75 +8 1872 1,760,168 41,672 23,434 26,518 1,851,792 +1 +1 Seine Year Population Growth since previous census (%) Catholics Protestants Jews Other religion, not reported or no religion Total population Reformed Church Other Protestants Reformed Church Total population 1861 1,911,186 9,584 16,620 15,196 1,074 1,953,660 1866 2,037,951 19,904 26,687 21,767 5,987 2,112,296 +108 +8 1872 2,118,652 22,217 24,831 24,319 30,041 2,220,060 +12 +5 (a) The Seine disappeared on 1 January 1968 when it was divided into four separate départements: Paris, Hauts-de-Seine, Seine-Saint-Denis and Val-de-Marne. Data at département level were not published in 1851. Source: Censuses, Statistique générale de la France.

Religious affiliation in Paris and in the Seine département^(a) according to the censuses of 1851 to 1872

41We cannot rule out a large-scale influx of Protestants coming from the provinces. It is decidedly difficult to discard Weber’s explanation whereby Protestants, much more so than Catholics, were attracted by the lure of city life and capitalism, most exemplified by Paris at that time. But the scale of subsequent change shows that this interpretation is not sufficient in itself. [16]

42While repression of Protestants was strong throughout France during the previous centuries, it was especially so in Paris. The specific effects of residing in the capital need to be examined in relation to the reduction in under-reporting. Families with a long history of residence in Paris had certainly not forgotten the Saint Bartholomew’s day massacre, which began in Paris, and the severe restrictions imposed upon their religion: Protestant churches were banned in the capital, and under the Edict of Nantes, Reformed Church members had to travel to Charenton to worship. For Reformed Church members arriving from rural areas, their residence in the capital may have been associated with a certain wariness. Indeed, Paris symbolizes the power of the state, a power which ordained the persecution of their ancestors. Under such circumstances, a more enduring reluctance to publicly divulge their religious affiliation is quite understandable. The reduction in under-reporting observed elsewhere in the 1850s did not occur in Paris until the 1860s.

43The trend appears to have levelled off by 1866. The reported Protestant population had barely risen in the 1872 census and, more importantly, its 1% growth rate was the same as that of the general population. For the Seine département, the increase (+12%) was only slightly higher than that of the département population as whole (+5%). Henceforth, Protestants were no longer wary of reporting themselves as such, even those living close to the capital. Unfortunately for religious statistics, the state chose this moment to stop counting their numbers.

Conclusion

44Did the administration make discretionary adjustments to offset the under-reporting of Reformed Church members in 1851? Though formal proof will probably never be found, the clues are troubling and consistent. The data must therefore by analysed with great caution. In his Mémoires, the baron Haussmann, at the time sub-prefect of the Haute-Loire département, points out that his Protestant brethren came to church in 1832 with a gun, which they left at the entrance, and that they were still “profoundly aggrieved by the “dragonnades” […], and the religious persecutions which marked the Bourbon restoration in 1815” (Haussmann, 2000, p. 119). Against such a backdrop, and even if these events had lapsed from memory, can we seriously imagine that these Reformed Protestants would unanimously report their religious affiliation to the census agent just twenty years later?

45Indeed, the Protestants’ favourite texts were able to calm any scruples about misrepresenting the truth – thus illustrating the particular form of individualism so often attributed to their kind. J.-P. Willaime (1992), for example, cites one of the Protestants’ much-loved references: “And when thou prayest, thou shalt not be as the hypocrites are: for they love to pray standing in the synagogues and in the corners of the streets, that they may be seen of men. […] But thou, when thou prayest, enter into thy closet, and when thou hast shut thy door, pray to thy Father which is in secret” (Matthew 6, 5-6). Moreover, the Old Testament contains passages which condemn the very principle of enumeration (2 Samuel 24, or 1 Chronicles 21) in which David’s decision to count the tribes of Israel was seen as a lack of faith in God (Willaime, 1992).

46It is doubtful, however, that these texts – which also feature in the Catholic corpus of course – were the decisive cause of under-reporting. Clearly, it is the Protestants’ status as a long-oppressed minority which appears to play the key role here. A similar process is at work today, resulting in under-estimation of Muslims in modern French surveys – for similar if not identical reasons (Dargent, 2003). Be it for nineteenth-century Protestants or for twenty-first century Muslims, it would be an act of credulity to assume that the data published by statistical sources can be “taken at face value”.