1Excess mortality in European cities in the nineteenth century, before the development of appropriate health infrastructures, is explained by poor living conditions and inadequate urban sanitation. However, the indicators measured are not always an accurate reflection of urban mortality conditions. First, they take account of recent immigrants whose risk of dying may be different from that of the local population. In particular, they may have moved to the city to receive hospital treatment. Second, mortality measures may be biased if the deaths and the populations at risk are not defined in a consistent manner. Deaths used in the numerator to calculate mortality rates are sometimes counted by place of death, while population estimates in the denominator are based on the place of habitual residence. Tina Van Rossem, Patrick Deboosere and Isabelle Devos examine the impact of this inconsistency by recalculating mortality levels in three Belgian cities in the early twentieth century on the basis of the habitual residence of the deceased.
2When reconstructing life expectancies and age-specific mortality estimates, demographers are confronted with the problem of unregistered migrations. The frequent misreporting or non-reporting of temporary attendees and absentees implies that some people die in a municipality other than their own. This makes correct calculation of the number of deaths and the population at risk much more complex, which in turn causes errors in the mortality rates of territorial areas and their relative health risks (Alter et al., 2009; Ocaña-Riola et al., 2009).
3Alongside illegal migration, short-term movements complicate the calculation of correct death rates (Skeldon, 1987). Students, soldiers, prisoners and hospital patients, for example, become part of a certain municipality for a couple of weeks or months, while still being registered as legal inhabitants in the population register of another. Furthermore, commuting workers are partly exposed to health risks in a municipality other than the one where they reside. For these reasons, in countries such as nineteenth-century Belgium with a compulsory population register (Poulain and Herm, 2013), there can be large discrepancies between the de facto population, i.e. the factual population consisting of registered and unregistered inhabitants present at a certain moment in time, and the de jure population, i.e. the legal population of registered inhabitants regardless of where they are at any given moment. Moreover, because of a certain negligence in registering permanent migration movements (especially in the past), we need to consider a specific subgroup of the de facto population that should be legally registered, but is in fact not. More specifically, these are people who have their habitual residence in a municipality, but are registered elsewhere. If all permanent migrants are registered correctly, the de jure population and the population with habitual residence would be exactly the same. Scholars have recognized the complexity of these differences in population numbers when attempting to calculate and compare demographic measurements of contemporary populations of developed and developing countries (Ocaña-Riola et al., 2009; Skeldon, 1987; Van Hook and Bean, 1998) as well as historical populations (Eggerickx and Debuisson, 1990; Thorvaldsen, 2006).
4With regard to demographic estimates of historical populations, there is an interesting ongoing debate on the impact of medical institutions on urban death rates (Mooney et al., 1999; Ramiro Farinas, 2007; Revuelta Eugercios, 2011). During the nineteenth century, the number of deaths in hospitals and other medical institutions gradually increased in large cities such as London (Mooney et al., 1999), but a significant share of the people who died in these institutions were not officially registered in the city itself. Hence, due to the high death risks in such institutions, their presence could artificially increase the death rates of large cities. By using data on the number of deaths of outsiders in medical institutions in Toledo around 1877, Ramiro Farinas (2007), for instance, concludes that “[We must] reassess whether mortality in the cities was in reality as high as it is often represented”. He argues that “the cities suffered because of their success as dynamic centres and for that reason, poles of attraction for migrants”. After excluding deaths of outsiders occurring in institutions, differences in life expectancy between Toledo and the countryside noticeably diminished (Ramiro Farinas, 2007). Hence, researchers should consider how population and mortality data were produced when interpreting historical urban death rates.
5In this article, we build upon these studies by producing what we believe to be the most accurate estimation of the death rates for urban populations since the mid nineteenth century in Belgium. Reconstructing correct death rates requires two basic elements. First, the nominator and denominator of the death rates should refer to the same population. In general, however, age-specific Belgian death rates for the second half of the nineteenth and the beginning of the twentieth century are based on de facto mortality numbers derived from the register of vital events (Mouvement de la Population et de L’État Civil) and on the population numbers with habitual residence drawn from population censuses (designated as de jure in the census).  Second, the population at risk has to be selected. We argue that the death rates should reflect the mortality experiences of urban populations with habitual residence, i.e. the population fully exposed to the hazards of a specific city. Consequently, the typically used de facto mortality numbers must be adjusted. Outsiders dying in medical institutions should be excluded, as should other deceased temporary attendees, because they were not fully exposed to the hazards of the city. Deceased people with habitual residence in the city, on the contrary, need to be taken into account. Moreover, the deaths of people with habitual residence who accidentally died outside the city, (henceforth labelled as “temporary absentees”) must also be considered. This concerns both deaths of people who were officially registered and therefore belonged to the de jure population, and deaths of those who were part of the subgroup of the de facto population that should have been legally registered. Hence, a reconstruction of the death rates of the population with habitual residence should correct the biases created by the typically used death rates, which include people who died in a particular municipality even if they did not habitually reside there and exclude deaths of habitual residents that occurred elsewhere.
6As a case study for this article, we selected three large Belgian cities: Brussels, Liège and Schaarbeek. Due to their size, abundant data on temporary migrants are available for these cities. Moreover, in Brussels and Liège, there appeared to be a large inflow of temporary attendees, while Schaarbeek showed a slightly higher number of temporary absentees than attendees. These different migration contexts make for a very interesting comparison. We corrected the death rates of these cities for the year 1910 using unique data from the register of vital events on the number of deaths of people who died inside and outside their municipality where they habitually resided. Unfortunately, these data are not age-specific. We therefore applied a supplementary source for the age distribution, namely the individual death certificates of these cities.
7In addition to correcting the death rates, we examined whether the previously noted excess mortality in Brussels around the turn of the twentieth century might partly be explained by the use of de facto numbers of deaths. Earlier studies remarked that life expectancy at birth in Brussels was significantly below the national average: between 1846 and 1910 the difference amounted to more than 7 years. More specifically, life expectancy at birth in Brussels in 1910 was 41.3 years, while the national average was 51.3. Compared with other large cities such as Antwerp, Ghent and Liège, there were still differences of 5 years and more, while the differences between Brussels and its suburbs could amount to 9 years (Devos and Van Rossem, 2015; Eggericx, 2013; Eggerickx and Debuisson, 1990). However, it is plausible that part of Brussels’ disadvantage was due to high numbers of deaths of temporary attendees. As the capital city, Brussels attracted many (unregistered) migrants, commuting workers and servants. Moreover, several large hospitals and asylums were located there.
8This article consists of four sections. Before focusing on the correction of the death rates, we briefly discuss the urban and industrial background of the three selected Belgian cities in Section I. The sources used for the correction and the applied method are described in detail in Section II. Next, we deepen our understanding of the importance of temporary migration movements in these three cities in Section III. The following three questions are addressed: What were the proportions of temporary attendees and absentees? What were their ages? How many of them died in medical institutions? The impact of the deaths of temporary migrants on the death rates is then discussed in Section IV. We compare the mortality rates based on de facto deaths and population numbers of habitual residents with those based on both deaths and population numbers of habitual residents.
I – Setting of the study
9To demonstrate the possible bias of temporary migration movements with respect to mortality estimates of urban environments in Belgium, we correct the death rates of 1910 for three cities: Brussels, Liège and Schaarbeek. The capital city of Brussels is situated in the middle of the country, while Liège is in Wallonia, the southern part of Belgium. Schaarbeek is a suburb of Brussels (Figure 1).
10The last Belgian population census before the outbreak of the First World War was conducted in 1910. By then, death rates in Belgium were already much lower than in the mid-nineteenth century. Life expectancy at birth was 51.3 years in Belgium around 1910, compared to only 37.8 years in 1846. Most of the progress in large cities had occurred at the turn of the century. Despite these major improvements, urban health penalties still existed. Moreover, differences in death rates and life expectancies between cities were still very large. Between 1846 and 1910, the largest urban health penalty was consistently observed in Brussels (Devos and Van Rossem, 2015). In 1910, 45% of the de facto deaths in the capital occurred in a medical institution, versus 34% in 1850, 29% in 1870, 28% in 1890 and 39% in 1900 (Ville de Bruxelles, 1851-1911). Since we are interested here in the impact of medical institutions on the interpretation of urban death rates, this high proportion of hospital deaths in 1910 is extremely relevant. Moreover, the quality of the administrative data had gradually improved since the mid-nineteenth century due to the growing use of pre-printed forms and the development of clear guidelines for data collection, enumeration and processing. Systematic supervision and controls were progressively introduced as well (Preneel, 2010; Vrielinck, 2013).
Locations of the three cities under study
Locations of the three cities under study
11Cities were selected for the case study in order to obtain a large number of temporary attendees and absentees. Since most migrations took place towards or from cities, our correction method is also of greater importance for urban than for rural areas. The population numbers for Brussels (177,078) and Liège (167,521) were fairly similar in 1910, while those for Schaarbeek were somewhat lower (82,480). Brussels became the capital of Belgium in 1831. The city grew very rapidly during the nineteenth century, as did other large cities such as Liège. Population growth was mainly driven by internal migration flows. Along with the strong population increase in these city centres, nineteenth-century Belgium also underwent a large-scale suburbanization process. The expansion of Brussels’ suburbs was especially rapid, which explains why, by 1910, Schaarbeek was part of the urban hierarchy (Eggerickx, 2013; Eggerickx and Debuisson, 1990).
12In Brussels, many labourers were employed in small factories specialized in construction, printing, luxury and consumer products, timber and furniture production. As the capital city, Brussels also provided many administrative employment opportunities and housed a large number of financial institutions. In Schaarbeek, the construction industry was particularly important (Bogaert-Damin and Marechal, 1978; De Beule, 1994; Van Rossem et al., 2017). Liège, on the other hand, was part of a large industrial basin with thriving steel, glass and machine-building industries (Leboutte et al., 1998). The industrial activities in these cities attracted many commuting workers.  Belgium’s small size together with its efficient railway system and tram network made commuting easy. Weekly and seasonal commuting was common in Flanders, while daily commuting was more prevalent in Wallonia (Deprez and Vandenbroeke, 1989).
13Furthermore, the aim is to compare cities with a high inflow of temporary attendees, such as Brussels and Liège, with a town with high numbers of deaths of temporary absentees, such as Schaarbeek  (Table 1). The numerous medical institutions located in the large cities of Brussels and Liège attracted many outsiders (Table 2). Schaarbeek also housed some institutions, but they appeared to attract fewer non-residents.
II – Belgian population and mortality data
14In order to reconstruct mortality figures for Belgian municipalities, age-specific population and mortality data are needed. For large municipalities and cities, age-specific population data on habitual residents can be derived from population censuses from 1866 onwards. Starting from 1886, age-specific mortality data on de facto deaths can be collected from register-based government population statistics (Mouvement de la Population et de L’État Civil, often simply referred to as Le Mouvement). However, the fact that age-specific mortality on the one hand, and population numbers on the other, are obtained from different sources may create a possible distortion of the demographic measurements based on these data. We already noted that death rates can be improved by using population and numbers of deaths among inhabitants with habitual residence rather than a combination of population data on habitual residents and de facto mortality numbers. As a result, by excluding temporary attendees and including temporary absentees, our rates refer to the deaths of people living in the city as habitual residents, and as such included in the census in a certain municipality. This correction method is applied for the three Belgian cities. In addition to the more generally used data from the population censuses and Le Mouvement, individual death certificates are also used to accurately estimate the age at death of people who died outside their municipality of habitual residence.
1 – Population censuses
15The first Belgian population census was conducted in 1846. It was organized by the Ministry of Interior and Public Education, while the actual counting and processing of data was entrusted to the municipal authorities. Approximately every 10 years after that, a new census was organized,  and from the volume of 1866 onwards, the age-specific population numbers – published for all municipalities with 10,000 inhabitants or more – referred to the population with habitual residence rather than the de facto population (Preneel, 2010). In addition to these numbers, the censuses continued to mention the total size of the de facto population, i.e. the population present in a municipality at the date of the census.
16The main objective of the population census of 1910 was to determine the size and composition of the population with habitual residence, described in the census as the de jure population.  The habitual residence of a person was the place where he or she should have been legally registered.  Because people sometimes did not register in a new municipality when they changed address, a person’s legal and habitual residence could differ. The administrators used two different types of bulletins to distinguish between the de facto population and the population with habitual residence.  All the people who habitually resided in the same household were recorded on household bulletins (bulletins de ménage), even if they were temporarily absent. The instructions defined a household as a group of people living in the same dwelling and sharing a common life. Kinship ties were not a necessary condition. Domestic servants and workers living with their masters were thus part of the same household. Moreover, religious people living in a collective home or military personnel living in barracks also formed a household. People who resided outside their habitual residence on 31 December 1910 were recorded on special bulletins as well (bulletins spéciaux, personnels ou collectifs). Individual special bulletins were intended for individuals, while collective special bulletins were used specifically for establishments comprising a group of people temporarily living together (such as hostels, prisons, hospitals or boarding houses) and for people who lived in mobile residences (such as boatmen, nomads and showmen) (Statistique de la Belgique, 1910, pp. 27-33).
17The criterion of habitual residence implies that people not regularly exposed to the hazards of a certain municipality were excluded from its population count in the census. Hence, they are also excluded from the population at risk that was used to construct demographic measurements. More specifically, the instructions of the census of 1910 explicitly required that the following nine groups be excluded from the population with habitual residence (Statistique de la Belgique, 1910, p. 28):
- people travelling for work, health or leisure, together with workers who return at regular intervals to their actual households;
- people being treated in hospitals and other public and private health establishments;
- people living in asylums;
- young people studying in a municipality other than that of their household;
- children placed with (wet) nurses;
- people placed in penitentiaries, hostels for beggars, refugee accommodation and charitable schools;
- soldiers under arms;
- officers on temporary assignment away from their habitual residence;
- clerics who are temporarily absent from their habitual residence or the residence where they are assigned.
18People in these categories were to be counted with the population of the municipality from which they were temporarily absent. Exceptions were made only for people belonging to categories 3 to 7 who were not part of a household elsewhere in Belgium or abroad. Their habitual residence was considered to be in the municipality where they resided.
2 – Mouvement de la Population et de L’État Civil
19From 1886 onwards, the Belgian central government published comprehensive and age-specific mortality statistics for all Belgian municipalities in Le Mouvement de la Population et de l’Etat Civil (Statistique de la Belgique, 1841-1976). The first edition of Le Mouvement dates back to 1841, but was only kept up to date from the 1880s. Up until 1976, it provided a comprehensive inventory of births, deaths, migrations and marriages for each of the 2,583 municipalities, 41 districts and 9 provinces in Belgium. The information in Le Mouvement was derived from local population and civil registers, of which the municipal government was obliged to make a yearly summary on pre-printed forms (Preneel, 2010). The data on deaths were derived from municipal death certificates.
20The registers of Le Mouvement were divided into eleven modules, known as cadres, each containing different sets of information on the vital events of local populations in a certain year. Data on general deaths (non cause-specific) were recorded in four cadres. In Cadre I the total number of deaths of men and women with habitual residence who died inside and outside the municipality was noted.  These numbers were used, together with information on births and migration, to calculate the population increase or decrease since the previous year. It is stated that the numbers should not be confused with those given in Cadres III and IV, which refer to the total numbers of deaths in the municipality (de facto). Cadre III gives a general overview of the total number of de facto deaths by sex, while Cadre IV breaks down these sex-specific mortality numbers by month. In Cadre V, finally, these de facto mortality numbers are broken down by sex and age. Stillbirths and infants who died before registration were never included in these numbers, but mentioned in separate columns.
21Thus, unlike the population numbers, the frequently used de facto data on age-specific mortality includes deaths of people who were only temporary exposed to the hazards of a certain municipality. As we will show in sub-section III.3, in large cities a very large share of these people were hospital patients. They also included commuting workers, travellers, occasional visitors, children placed with wet nurses, etc. On the other hand, people who died while temporarily absent from the municipality where they habitually resided are not included. Luckily, we have data on the number of inhabitants with habitual residence who died inside and outside the municipality from Cadre I of Le Mouvement. This gives us unique information on the total number of deaths of people habitually exposed to the hazards of a certain municipality. We used death certificates to estimate their age distribution. As a result, we are able to explore in detail how deaths of people outside their municipality of habitual residence affect the comparability of Belgian urban mortality figures.
3 – Municipal death certificates
22Municipal death certificates are used in this article to estimate the age distribution of deaths of temporary attendees and absentees in the three Belgian cities. From June 1797 onwards, the “Belgian” départements created under French rule were bound by the decree on official civil status registration introduced by the French government in 1792.  Civil servants were required to complete a certificate for each birth, marriage and death that took place in their municipality. These certificates were kept in three different registers. For greater homogeneity and uniformity, standard versions of the certificates were introduced around 1830. These death certificates gave detailed information on the deceased: date of death, name, occupation, age, municipality of birth, place of death, legal and habitual place of residence and whether they were widowed, married, divorced or single. The name of their spouses and of their parents was also indicated, and if they were still alive, their occupation. Sometimes, the address of these family members was also recorded. Finally, they also gave the name, occupation and municipality of the two witnesses required to sign the certificate (Bracke, 2008; Vanhaute, 2003).
23If someone usually lived outside the municipality where he or she died, the civil registration of the municipality of habitual residence had to be informed. A duplicate of the death certificate was issued by the receiving municipality, and the deceased person was erased from the population registers. The duplicate certificate gave the same information about the deceased as the original. However, because not every municipality used pre-printed forms, details were sometimes omitted. Furthermore, there was also a frequent time lag between the death of a person outside his or her habitual residence and the declaration thereof. Especially when a death occurred in a foreign country, the certificates reached the relevant civil registration only after several months or even years. Moreover, errors or omissions in address changes of the deceased also occasionally complicated and delayed the administrative process.
III – Death outside the municipality of habitual residence
24Before attempting to estimate correct death rates, we will examine the scale of temporary migration movements in the three Belgian cities around 1910, the number of deaths of temporary absentees and attendees, their age structure and the number of deaths occurring in medical institutions.
1 – The number of deaths
25Table 1 shows, for the selected Belgian cities, how many habitual residents died in 1910 while absent from the city, as well as how many temporary attendees died there. According to Le Mouvement, the proportion of deaths of temporary attendees and absentees was especially high in Brussels. The number of deaths of male temporary attendees in Brussels, for instance, was equivalent to 28% of the deaths of habitual residents, while the proportion of deaths of temporary absentees was equivalent to nearly 17%. The percentages for females were somewhat lower: 25% for attendees and 16% for absentees. In Liège, the numbers were much more moderate, but the pattern is the same: a larger number of deaths of temporary attendees in proportion to habitual residents in the city (17% for males and 10% for females) than of deaths of temporary absentees occurring elsewhere (9% and 5%). In the Brussels’ suburb of Schaarbeek, on the other hand, the opposite was true: the proportion of males and females habitually residing in Schaarbeek and dying elsewhere (17% and 14%) was slightly higher than the number of deaths of temporary attendees dying in the town (15% and 12%). This clearly reflects a difference in temporary migration to Brussels and Liège on the one hand, and to Schaarbeek on the other. Furthermore, these differences imply a decrease in the number of deaths for Brussels and Liège and an increase for Schaarbeek if deaths of habitual residents are used rather than de facto deaths.
Mortality of males and females in Brussels, Liège and Schaarbeek by place of registration and death, according to Le Mouvement, 1910
Mortality of males and females in Brussels, Liège and Schaarbeek by place of registration and death, according to Le Mouvement, 1910Note: (a) number of deaths of people with habitual residence who died inside the city; (b) number of deaths of people with habitual residence who died outside the city; (c) number of deaths of people without habitual residence who died inside the city; (d) deaths of temporary absentees as a percentage of the deaths of habitual residents who died inside in the city; (e) deaths of temporary attendees as a percentage of the deaths of habitual residents who died inside in the city; (f) total number of deaths of people with habitual residence; (g) total number of deaths of people who died inside the city.
2 – Age at death of temporary attendees and absentees
26The age distribution of the deceased temporary attendees and absentees in the towns and cities under study (Figures 2 to 7) was derived from municipal certificates. It is important to note that there are some discrepancies between the number of deaths of temporary attendees and absentees noted in Le Mouvement and the number of certificates we were able to trace back (see Appendix Table A.1). We decided to use the death certificates solely for information on age distribution and to apply the usually higher numbers of Le Mouvement in our analysis (Section IV). As Le Mouvement was updated in the subsequent year, we assume that information obtained later was included nonetheless and that the numbers were thoroughly verified by the statistical office.
27For Brussels, we immediately note that a large proportion of the deceased temporary attendees and absentees were infants and young children (Figures 2A and 2B). The presence of the Hospice des Enfants Assistés (Institute for the Protection of Children) in Brussels largely explains the high numbers of deaths of children from outside the city; 50% of male deaths and 39% of female deaths of temporary attendees below the age of 5 occurred in this Institute. The large number of deaths of children with habitual residence dying outside Brussels may have been due to the practice of entrusting infants and children to wet nurses in neighbouring municipalities. An estimated 48% of the deceased absent boys and 57% of the deceased absent girls below age 5 died in one of Brussels’ suburbs. Although wet nursing was much less common in Belgium than in France, it was still very widespread around the turn of the twentieth century. By examining the reports of the provincial medical commissions, Debuisson (2001) found that at the end of the nineteenth century in cities such as Brussels and Antwerp, many children of working women were fed by wet nurses. Most wet nurses in Belgium lived very close to the mother’s place of work or residence (Masuy-Stroobant, 1983). The proportions of temporary attendees and absentees in Brussels are quite low at ages 5-15, but rise again for adult age groups, reaching high levels for the age groups between 20 and 60 years. Two determinants are largely responsible for these high proportions: first, transportation towards hospitals (Sub-section III.3) – the large hospitals Hôpital Saint Jean and Hôpital Saint Pierre were located in the city centre; and second, the large number of commuters in Belgium at the turn of the twentieth century due to a very dense railway network and a well-developed system of daily and weekly rail passes (Footnote 2) (Deprez and Vandenbroeke, 1989; Pasleau, 1995). The proportions are somewhat smaller at older ages, although the proportion of female temporary absentees remains quite high.
28The proportions of temporary attendees and absentees belonging to the youngest age groups were smaller in Liège (Figures 2C and 2D) than in Brussels. Those of older children were equally low. Hence, most temporary attendees and absentees in Liège were adults and elderly people. Generally speaking, the largest proportions of male temporary attendees were in the younger adult age groups, while the largest proportions of temporary absentees were found in the age groups above 50 years. Female temporary attendees and absentees in Liège were distributed more or less equally between the different adult and elderly age groups, with a peak in the age categories between 55 and 65 years. As for Brussels, the high proportion of temporary migrants belonging to the adult age groups seems to have been strongly related to hospital admission and to commuting. Many people in the area around Liège commuted to work in the flourishing mining and metal industry (Pasleau, 1995). The large proportions of temporary absentees at older ages are puzzling. Only a few cases can be attributed to admission to a medical institution outside the city or a return to one’s place of birth. Hence, there must have been other movements on a larger scale.
Age structure of the deceased temporary absentees and attendees in Brussels, Liège and Schaarbeek by sex, according to the death certificates, 1910
Age structure of the deceased temporary absentees and attendees in Brussels, Liège and Schaarbeek by sex, according to the death certificates, 1910Note: Except for deaths at ages 0 and [1;5), age x means death at age [x;x+5). 85 signifies 85 and over.
29Finally, Figures 2E and 2F show that the proportion of young and very young temporary attendees and absentees in Schaarbeek was quite moderate compared to Brussels and Liège. However, the proportion of deaths of female infants and young girls temporarily present in the town – most of whom had their habitual residence in one of Brussels’ other suburbs – was even larger than in Brussels. Hence, a share of these deaths probably concerned infants placed with wet nurses in Schaarbeek. As in the other studied cities, only a few temporary attendees and absentees were aged between 5 and 15 years. The proportions of adults vary greatly by sex and specific age group. These large fluctuations are, of course, strongly linked to the low numbers of deaths obtained from the certificates (Appendix Table A.1).
30In sum, three findings can be derived from these age distributions. First, most temporary absentees and attendees were adults aged between 20 and 60 years; second, compared to the temporary attendees, significant proportions of the absentees were older than 60 years; and third, the age-specific proportions differed significantly by sex and city. The large differences in the proportions of temporarily present and absent infants and young children are especially notable. Infant and child mortality significantly influence the calculation of life expectancies and mortality risks. Hence, when correcting the Belgian cities’ death rates, it is vitally important to take into account the specific age distribution of the groups of deceased temporary attendees and absentees in every studied city.
3 – The impact of medical institutions
31Besides providing data on the age structure, death certificates also give information on some other characteristics of the temporary migration process. Using information on the place of death and the occupation of the witnesses, we estimated how many temporary attendees and absentees died in medical institutions. The results are presented in Table 2. As “deaths occurring in a medical institution” are recorded in various ways on death certificates, we used different methods to count them in each of the cities studied.
32For the groups of temporary absentees, we relied for all three cities on the occupations of the witnesses or on the mention of a medical institution as the place of death in the duplicate certificates. Because we assume that not every municipal officer added the available non-mandatory information on the medical facilities concerned, it is plausible that we underestimated the impact of medical institutions for this group. The noted percentages for the group of temporary absentees should thus be considered as a minimum.
Proportion of deaths in medical institutions of temporary absentees and attendees in Brussels, Liège and Schaarbeek by sex, 1910
Proportion of deaths in medical institutions of temporary absentees and attendees in Brussels, Liège and Schaarbeek by sex, 1910Note: The medical institutions included in this table are hospitals, hospices, institutions for the protection of children, asylums and sanatoria.
(a) number of deaths of people with habitual residence who died outside the city; (b) number of deaths of people without habitual residence who died in the city.
33Our reconstruction of the group of temporary attendees dying in medical institutions in Brussels and Schaarbeek is based on the occupation of the witnesses or on the address of the place of death. On most of the death certificates of people who died in medical institutions, the first witness is the director of that hospital or hospice. In addition, we also used the Brussels Almanacs of Trade and Industry to check that a medical institution was located at several addresses where a number of deaths occurred in 1910 (Administration communale de Bruxelles, 1910). For Liège, we could only rely on the address of the place of death. Hospital directors were not mentioned as witnesses there. We used the current addresses of medical institutions together with the frequency of deaths occurring at certain addresses to deduce the location of large medical institutions in Liège around 1910. 
34The estimated percentages in Table 2 show that medical institutions were of extreme importance in the temporary migration process. Seemingly, many people were temporarily present in the large cities of Brussels and Liège for medical reasons. In Liège more than 60% of the deceased male and female temporary attendees died in a hospital or other institution, and in Brussels more than 80%. In the smaller town of Schaarbeek, the percentages were much lower: 22% for females and 39% for males. The presence of large hospitals thus seems to have artificially increased the de facto death rates in large Belgian cities. Our results furthermore suggest that only a small percentage of the deceased temporary absentees from Brussels and Liège died in a medical institution elsewhere. For Schaarbeek, these percentages are significantly higher, ranging between 26% and 30%. Approximately 18% of absentee deaths of habitual residents of Schaarbeek occurred in an institution in Brussels.
35In addition to information on medical institutions, the death certificates and duplicate certificates also provide some scattered but unique information on other places of death. For instance, several certificates mention that children died in the house of their foster mother, which confirms our hypothesis about temporary placement of children with wet nurses. Deaths of adolescents and adults are also recorded in the houses of their (grand) parents, in their birth town or in factories. Other detailed information concerns deaths occurring in a railway station or in the person’s country house. These anecdotal references demonstrate that despite the large impact of medical institutions, there were also many other very diverse motives behind temporary movements.
IV – Differences between the original and corrected death rates
36The original age-specific death rates of the three cities, based on de facto deaths and the population with habitual residence, are presented in Appendix Table A.2 together with the corrected rates based on deaths of people with habitual residence. Since the number of de facto deaths in Brussels and Liège was higher than the number of deaths of habitual residents, the general corrected death rates are somewhat lower for men and women. For most age groups, the differences are minor, however. Life expectancy at birth changes in Brussels from 37.3 years to 39.9 years for males and from 45.3 years to 47.7 years for females when the corrected rates are used. For Liège, life expectancies rise from 46.2 to 48.0 years for males and from 51.8 to 52.6 years for females. The difference with Brussels thus decreases slightly when the corrected numbers are used (Table 3). As the difference between the number of temporary absentees and attendees was very small in Schaarbeek, the corrections to the death rates of this town are even smaller, and the life expectancies at birth barely change: from 52.1 to 52.0 years for males and from 57.6 to 57.7 years for females. Nevertheless, most transformations of age-specific death rates in Schaarbeek show an increase due to slightly higher numbers of deaths of habitual residents than of de facto deaths. However, there are some exceptions. The corrected death rates for girls are, for instance, lower than the original rates (Appendix Table A.2). These contrasts are, of course, related to the different age compositions of temporary attendees and absentees. When the corrected life expectancy is used for Brussels, the difference with respect to Schaarbeek decreases by more than 2 years. Hence, the differences in measures of mortality and population numbers influence the interpretation of urban death rates and life expectancies.
Original (O) and corrected (C) life expectancies of residents of Brussels, Liège and Schaarbeek, by sex and difference with Brussels, 1910
Original (O) and corrected (C) life expectancies of residents of Brussels, Liège and Schaarbeek, by sex and difference with Brussels, 1910Note: The original life expectancies are based on de facto mortality and population with habitual residence (O), while the corrected life expectancies are based on mortality with habitual residence and population with habitual residence (C).
37For comparison of death rates, we are particularly interested in the possibility that excess mortality in Brussels in the early twentieth century was an artefact. Table 4 shows a comparison between the rates based on de facto deaths and those based on deaths of habitual residents. The death rates of Brussels are the reference; numbers below 100 in Liège or Schaarbeek reflect excess mortality in the capital city for that specific age group. The lower the number, the greater Brussels’ disadvantage. The figures in bold indicate cases where excess mortality declines when the rates are based on deaths of habitual residents.
38The results in Table 4 clearly show that most age groups in Brussels experienced large health penalties compared to the other two cities. Large differences are observed, for adults especially. For instance, the de facto death rate of men between 30 and 35 years in Liège was only 70% of that of Brussels, and in Schaarbeek just 50%. The percentages for women are 71% and 43%. The change in rates when deaths of habitual residents are used modifies the comparison. For several age groups, Brussels’ excess mortality drops significantly, and for some it even disappears completely due to the exclusion of deaths of temporary attendees and the inclusion of deaths of temporary absentees. Looking again at the age group between 30 and 35 years, the percentages for men change to 67% in Liège and 56% in Schaarbeek and those for women to 100% and 80%. For men, Brussels’ excess mortality with respect to Liège increases slightly. For women, on the other hand, it disappears completely. Compared to Schaarbeek, the disadvantage of men and women between ages 30 and 35 becomes significantly smaller. Nevertheless, even after correction, differences with the other cities still remain very large. The rates of habitual residents in Liège and Schaarbeek are frequently only 50%, 60% or 70% of those of Brussels (see Table 4).
Age-specific original (O) and corrected (C) death rates (per 1,000) of male and female residents in Liège and Schaarbeek compared to the death rate of Brussels, 1910
Age-specific original (O) and corrected (C) death rates (per 1,000) of male and female residents in Liège and Schaarbeek compared to the death rate of Brussels, 1910Note: A decline of excess mortality in Brussels is shown in bold.
V – Conclusion
39The calculations in this article show that the interpretation of death rates and life expectancies should take into account the way population and mortality data are produced. Because of temporary migrations, the number of deaths may differ substantially according to the legal, factual, or habitual residence of the deceased. Two crucial factors need to be considered. First, the nominator and denominator of the death rates should, in principle, refer to the same population at risk. Second, the population at risk needs to be chosen. We argue that the most accurate death rates are those that reflect the mortality experiences of the population with habitual residence, since it is these residents who are fully exposed to the hazards of a certain municipality. Hence, temporary attendees should be excluded from the calculations and temporary absentees included. This eliminates selection effects caused by the presence of hospitals or other institutions in large cities.
40The merits of using death rates of the population with habitual residence were shown by examining the 1910 death rates for three large Belgian cities. We used data from population censuses, the register of vital events (Mouvement de la Population et de L’État Civil) and individual death certificates. Considering that census data cover people with habitual residence, while vital statistics use de facto deaths, we started our analysis with an overview of the size and composition of the group of deaths of temporary attendees and absentees in the selected cities of Brussels, Liège and Schaarbeek. In the large cities of Brussels and Liège, the number of deceased temporary attendees was higher than that of temporary absentees. In the Brussels’ suburb of Schaarbeek, the number of deaths of temporary absentees was slightly higher than that of attendees. Most of the deceased temporary migrants were adults or elderly people, although large proportions of infants and children were also noted. The age-specific distributions clearly differed by city and sex. Furthermore, the results indicate that the majority of temporary attendees in Brussels and Liège died in a medical institution. After correction, the death rates in Brussels and Liège decreased slightly for several age groups, while for most age groups in Schaarbeek there was a slight increase. Clearly, when comparing the death rates of urban populations, it is important to take their habitual residence into account. In fact, our results show that the well-known excess mortality in Brussels at the beginning of the twentieth century was partly caused by the high numbers of deaths of temporary attendees in the city centre. Life expectancy at birth increases from 41.2 to 43.8 years when the correction is applied. Still, even with this correction, the death rates of most age groups are still much higher in Brussels than in the other cities.
41In sum, it seems that the presence of large medical institutions in particular can strongly distort the calculation and interpretation of urban death rates by artificially increasing rates in large cities. Hence, we agree with Ramiro Farinas (2007) that the living conditions in many large cities during the nineteenth and early twentieth century were probably less lethal than generally assumed. Furthermore, our research also shows that for contemporary populations likewise, it is important to use correct measures of population numbers and mortality when calculating and comparing death rates.
AcknowledgementsThis research was funded by FWO, The Research Foundation – Flanders. We thank the editors and the three anonymous reviewers for their comments on our manuscript and their excellent suggestions for improving the article.
Numbers of deaths of male and female temporary attendees and absentees in Brussels, Liège and Schaarbeek, 1910 according to Le Mouvement (MV) and death certificates
Numbers of deaths of male and female temporary attendees and absentees in Brussels, Liège and Schaarbeek, 1910 according to Le Mouvement (MV) and death certificatesNote: (a) number of deaths of people with habitual residence who died outside the city; (b) number of deaths of people without habitual residence who died inside the city.
Age-specific original (O) and corrected (C) death rates (per 1,000) of male and female residents in Brussels, Liège and Schaarbeek, 1910
Age-specific original (O) and corrected (C) death rates (per 1,000) of male and female residents in Brussels, Liège and Schaarbeek, 1910
From 1866 onwards, the Belgian population censuses already used a correction of the de jure population by including people not legally registered in the population registers of a municipality even though they habitually resided there (Statistique de la Belgique, 1910).
Data from the industrial census of 1896 provided information on the proportion of commuting workers in the three cities. The census delivers, per municipality, the total number of resident workers employed in the municipality or in a neighbouring one as well as the total number of workers in employment. The results indicate that in Brussels at least 23% of male workers and 7% of female workers commuted to the city. In Liège these percentages were 14% and 6%, and in Schaarbeek 27% and 12% (Statistique de la Belgique, 1896).
The significant number of habitual residents of Schaarbeek who died outside the town was probably related to the proximity of Brussels. Large medical institutions were located in the capital.
During the nineteenth and early twentieth century, Belgian population censuses were conducted on 31 December 1856, 1866, 1880, 1890, 1900 and 1910.
“[La population de droit] est basée sur la résidence habituelle. La population de droit d’une localité est donc composée des personnes qui ont dans cette localité leur résidence habituelle, qu’elles y soient ou non présentes le jour du recensement. La résidence habituelle se détermine d’après les règles établies en matière d’inscription aux registres de population.” (Statistique de la Belgique, 1910, p. 2). [[The de jure population] is based on habitual residence. The de jure population of a locality is thus composed of persons whose habitual residence is in this locality, whether or not they are present on the census day. Habitual residence is determined on the basis of the rules established for registration in the population registers.]
“La résidence habituelle d’une personne est là où, d’après les instructions en vigueur, elle doit être inscrite au registre de population.” (Statistique de la Belgique, 1910, p. 2) [A person’s habitual residence is the place where, in compliance with current instructions, he or she must be registered in the population register.]
The difference between the de facto population and the population with habitual residence was especially large in big cities. For instance, the de facto population of Brussels on 31 December 1910 was 180,943 while that of people with habitual residence was 177,078 (Statistique de la Belgique, 1910).
In Le Mouvement, these deaths are designated as “de jure deaths”. However, by comparing these numbers with those from the death certificates, we determined that they are deaths of people with habitual residence.
Before the introduction of official civil status registration in Belgium, parish priests had already kept registers of baptisms and burials for several centuries. Registration of baptisms was mandatory in Catholic regions after the Council of Trent (1563); the registration of burials was made compulsory by the Rituale Romanum (1614) (Devos and Vandenbroeke, 2004).
The following addresses in Liège were considered as housing a medical institution: Boulevard de la Constitution 66; Boulevard de la Constitution 75; Rue des Wallons 74; Rue Montagne Sainte-Walburge 24. The proportion of deaths in medical institutions is probably slightly underestimated because we could not trace back with certainty the presence of a medical institution for several other addresses where more than two deaths occurred in 1910.