Introduction

1In Western societies, the demographic revolution of the nineteenth and early twentieth century removed populations from the secular paradigm of high mortality and high fertility: a demographic mechanism only able to guarantee uncertain growth in the long run. In some decades, at different speeds and in different ways, life expectancy increased dramatically, passing from around 30 to over 80 years in most Western countries. Later, couples started, more or less consciously, to reduce family size to the point of halving fertility levels within two generations. However, the drop was so dramatic that since some decades, births are no longer able to ensure population stability.

2For Italy, as for the rest of European countries, the process of fertility decline has been carefully studied, at least in terms of timetable and geography, thanks to the work and results of the European Fertility Project. [1] The volume by Massimo Livi Bacci, published in 1977 within the national monographs series of the Princeton Project, remains a reference work for reconstructing fertility decline in Italy. [2] A process that occurred with different rhythms and at a different pace across regions and social strata. In fact, fertility control started later and spread more slowly in the South as well as in Sardinia and Sicily. In the same way, signs of fertility decline were first noted in some social-group forerunners of fertility control: these groups included nobles, Jews, and the urban élite (Livi Bacci, 1977, 1986).

3The general framework of Italian fertility decline has been established. However, there are still elements and mechanisms that need to be investigated more carefully. For instance, the start of the process of fertility decline seems to have occurred earlier than had been thought. A tradition of studies, in fact, has placed the beginning of that decline just after Italian unification (1861): the trend becomes evident in some of the first statistics produced by the new Office of Statistics (DIRSTAT) of the Italian Kingdom. Some recent studies concerning Northern Italian populations provide, however, evidence that places the onset of mortality decline at the beginning of the nineteenth century, and the start of fertility decline well before Italian unification (Breschi et al., 1999; Fornasin et al., 2012; Rossi, 1994). Even more uncertain is the way that behavior and birth control methods spread among families and people. Decades of aggregate-level researches on historical fertility decline in Western countries have revealed its connections to industrialization and urbanization (e.g., Brown, Guinnane, 2002; Carlsson, 1966; Davis, 1945; Dribe, 2009; Galloway, Hammel, Lee, 1994; Notestein, 1945), modernity and education (Casterline, 2001; Cleland, Wilson, 1987; Knodel, van de Walle, 1979; Lesthaeghe, 1977, 1980; Lesthaeghe, Wilson, 1986), and the prior decline in infant and child mortality (Davis, 1945; Dyson, 2010; Easterlin, 1996; Haines, 1998, Notestein, 1945; Oris, 1995; Reher, 1999; Reher, Sanz-Gimeno, 2007). Despite the importance of this evidence and a number of theories focusing on the greater relevance of education, women’s relative wages and independence, new attitudes and norms, and secularization (Guinnane, 2011), the mechanisms producing the demographic revolution of the nineteenth and early twentieth century are still unclear. Today, there is keener sense that only a micro-level approach will allow researchers to get a proper understanding of the process of fertility decline. Micro-studies have, though, been little used in Italy. Indeed, it is only in the last years that historical demographers have begun to cross the classical divide of Italian unification: which has represented, sometimes unwittingly, the border between historical demography and demography tout court (Del Panta, Rettaroli, 1994, 6). On the other hand, the first ad-hoc analyses of the reproductive behavior of Italians date back to the 1970s. [3] This combination of historical demographers working prior to 1861 and demographers proper concentrating on the Post War period has created a grey area just over the central period of the fertility decline. [4]

4In order to shed some light on that period, in the last years several research projects were launched with the intention of reconstructing, at the individual level, the demographic history of some Italian populations and the changes in demographic behavior across the transitional period (Pozzi, Sonnino, 2012). At the same time, some studies began to use individual level data and information on the reproductive history of Italian women collected on national censuses in the twentieth century. In Italy, the first survey on fertility was carried out with the national census of April, 21 1931 (ISTAT, 1936). A second survey was carried out thirty years later, with the census of October, 15 1961 (ISTAT, 1974). Just as in 1931, information concerned all ever-married women, respectively 8 and 13.7 million. This mass of documentary material was used by the national institute of statistics to compute only aggregate measures, such as rates, means, and other indicators. [5] Census returns were processed at the municipal level so that the original material can be found only in city halls. This explains why it is so difficult to extend the study to large cities, [6] although the recovery and computerization of the original census forms seems promising for the study of the evolution of Italian reproductive behavior. In particular, the information recorded on the occasion of the special surveys on the reproductive history of women (age at marriage, offspring size, etc.) could be integrated with information usually collected in censuses, such as the socio-demographic conditions of women and their families.

5The present work continues in this tradition of research, and, to some extent, it represents an in-depth examination and analysis of previous studies carried out on the original material of the 1961 national census. Thanks to those data we have the opportunity of investigating the reproductive behavior of Italian women born from the 1870s to 1925, those who led recovery in fertility after WWII and the baby boom of the 1960s.

6Our starting points are the “Census returns of the 10^th general census of the population – 15 October 1961” for ten communities located in very different areas of Italy: these communities have different geography, social structure, and demographic characteristics. In some previous papers on the Italian Census of 1961, [7] we have investigated fertility decline in Italy based on the reproductive histories of ever-married women that the survey exceptionally collected. Those articles shed some light on the key role that female education played on fertility, especially in the first stages of the demographic transition. Our analysis was based on four case-studies – Turriaco, Novellara, Casalguidi, and Alghero. The reproductive histories of these communities were used to trace the changes in the reproductive pattern during that exceptional period of Italian demographic history. In the present paper, we intend to take a step forward by improving some of the limits of the previous article. First, we have enlarged our dataset to include six more populations. The issue of representativeness is key for Italy given its great variability in demographic systems and patterns of demographic transition. Differences were, in fact, present not only between regions but also within the same region. The coastal town of Alghero, for instance, was not representative of the whole of Sardinia: inland regions had different characteristics. Likewise, Turriaco was not representative of the mountain area of Friuli, which accounts for a large part of the region. Moreover, Southern Italy, with its characteristic demographic system and culture, was not incorporated at all in the previous papers. To give more comprehensive coverage we include here four new communities for Sardinia (Austis and Seulo, Macomer and Porto Torres), another one from the mountainous area of Friuli (Lauco) and a community in Campania (Massa Lubrense), a municipality situated in the province of Naples. Second, in the previous papers the analysis was focused on completed-fertility women. This was done in order to conform our work to the analyses done by ISTAT on the census of 1961, but it limited any research to women born prior to 1912. This limitation prevented us, however, from taking fully advantage of the potentialities of the data. In particular, the analysis did not consider women with incomplete fertility, who were the protagonists of the last and determinant stage of fertility decline. Women born after 1911 were, in fact, likely to be better educated, and, therefore, more open to innovation and the ideas of modernization (Cleland, 2001; Van de Putte, 2007), such as more awareness of fertility control and use of innovative methods for the same (Cleland and Jejeebhoy, 1996). Moreover, these women were more likely to participate in the labor force thereby delaying their marriages (United Nations, 1995; Chaudhury, 1984). In short, the younger the woman in the census of 1961, the more she was likely to have independence and autonomy within the family, an element that was positively associated with women’s education (Dyson, Moore, 1983; Basu, 2002). Using the 1961 general census at the individual level gives us an opportunity to analyze the complex interplay between socioeconomic status and fertility. The literature contains a widely-held view that the association between social status and fertility changed in conjunction with the fertility transition. However, as recently pointed out by Dribe and colleagues (2014), in a review of the theoretical discussions on the role of socioeconomic status on fertility during the transition, this is still a controversial issue. It requires more solid evidence and further in-depth micro-analytic studies, something we hope to achieve here.

7The present paper is organized into five sections. The first section is the introduction. In the second section, after a description of the main characteristics of the 1961 census returns, an overview of some of the most important ISTAT elaborations are given. The third section describes the socioeconomic characteristics of the communities as they emerged from the 1961 census, a snapshot of a country in the middle of its transformation from an agrarian to industrial economy. The fourth section provides a descriptive analysis of the fertility of ever-married women in our studied populations and a comparison, also at a regional level, with the results and elaborations produced by ISTAT. Finally, the fifth section contains the micro-level analysis of the reproductive histories of ever-married women in the ten communities.

The 1961 Fertility Survey

The source

8The information on the reproductive history of Italian women collected in the 1961 general census are described in section VI of the census form, “Information on the number of children-ever-born to married, widowed, divorced, or separated women legally belonging to the family” (see fig. 1). The section has two parts. Part A, called “Information on marriage”, takes into consideration each ever-married woman living in the family, reporting the date (month and year) of her most recent/current marriage, the possible date (month and year) of widowhood (spouse’s date of death) or divorce/separation, the birth years of parents, and the dates (start and end) of possible previous marriages. Part B, called “Information on children from current and previous marriages”, records the number and sex of children ever born and the number and sex of those still alive at the time of the census (whether they lived with the parental family or not). The years of birth of all children are also recorded (without gender distinction) – which perhaps represents the greatest novelty of this survey with respect to that of 1931. This data can then be combined with information regarding each ever-married woman taken from other sections of the family sheet: this includes date and place of birth, education, and professional status and field. Alternatively the data can be related to characteristics of other members of the family group and/or the household’s living conditions: possessory title, number of rooms, presence of bathroom, availability of drinking water etc.

Fig. 1

Section VI of the 1961 census return. Reconstruction of the reproductive history of all ever-married women living in the family

Section VI of the 1961 census return. Reconstruction of the reproductive history of all ever-married women living in the family

9It is possible to reconstruct, at least approximately, the complete reproductive history (in terms of intensity and frequency) of the women (married at least once) born before 1911, and the incomplete history of those born afterward.

10As often happens with retrospective analysis, the Fertility Survey of 1961 also poses selection bias problems due to differential mortality and migration by fertility. The relationship between fertility and the mortality of women is multifaceted and the empirical evidence is – at least in part – contradictory for historic populations that practised only archaic forms of fertility control (Dribe, 2004; Doblhammer, Oeppen, 2003; Gagnon et al., 2009; Le Bourg, 2007; Westendorp, Kirkwood, 1998). Also uncertain among historic populations, is the relationship between fertility and maternal mortality because of the difficulty of establishing maternal mortality (Scalone, 2014; Manfredini et al., 2015). Between fertility and longevity, the empirical evidence would, meanwhile, seem to suggest more survival to advanced ages among women with fewer children and, mostly, higher age at last birth (Caselli et al., 2013). In more recent populations practicing more “modern” contraceptive techniques, the results have been clearer. The evidence for fertility and relative mortality risks suggests a U-shaped correlation between number of children and mortality risk (Dior et al., 2013; van Bavel, 2014): childless women have higher mortality risk than parous ones; women with few children have lower mortality, while women with many children (more than 5) experience higher mortality (Le Bourg, 2007), though the differences depend on context (Grundy, 2009; Doblhammer, 2000). This mortality differential would have caused a selection against high-fertility women in the census of 1961. This would, in turn, have resulted in an artificially lower completed family size for these groups. If we assume that non-educated and low-SES women were those with the highest fertility, any positive evidence of this differential between educational and/or SES groups can be considered a fortiori robust.

11The relationship between fertility and the migration of women, couples and families is an even more uncertain issue for historical populations: uncertain because of the limited availability of nominative sources concerning migration. Emigration was more likely to have involved childless women than high-parity women, as it has been demonstrated for contemporary Italy. However, we cannot rule out some over-estimation of fertility especially for the oldest cohorts (see Caltabiano, Dalla Zuanna (2015) for similar considerations on the Veneto region in the 1971 census fertility survey; see also Van Bavel (2014) on the Belgian census of 1981). Some studies prove, though, that this kind of bias is negligible (see Caltabiano, Dalla Zuanna (2015) for the fertility survey of 1971) In any case, these possible mis-estimations should not eliminate (as we shall see) the large differentials by level of education and SES. Finally, immigrant women do, at least according to our data, not appear to show any significant difference in marital fertility compared to native women.

ISTAT publications about the 1961 Fertility Survey

12The first statistics elaborated from the data of the 1961 Fertility Survey were produced by ISTAT only in 1974. This publication referred almost exclusively to “children born alive to women married only once before the age of 45 up to 1961”, a criterion that was adopted to simplify the data processing and number of tables (ISTAT, 1974, 9). Moreover, the analyses were conducted in a longitudinal perspective, using birth cohorts of ever-married women.

13Table 1 reports the number of children ever born (total and partial) to women married only once before the age of 45 by birth cohort. The data refer to the entire country and its geographical divisions. There are obvious limitations inherent in this type of comparison: figures are undoubtedly affected by variations in age at marriage and selection by mortality and migration, even for generations born before 1912. What emerges is the striking variability over time and across regions. As for changes over time, completed family size for whole Italy dropped from 4.4 to 3.3 children ever born per married woman (-25%) between the generations born before 1887 and those born between 1907 and 1911. This decrease was much more pronounced in North-Western Italy (-29.2%) and much less dramatic in Southern Italy (-9.1%), probably suggesting the more rapid diffusion of birth control practices in the North (Livi Bacci, 1977). This evidence is confirmed by the trend of the ratio of completed family sizes of South and North-Western Italy, which grows steadily, passing from 1.38 (4.96/3.60) for the generations born before 1887 to 1.77 (4.51/2.55) for the 19107-11 birth cohorts.

Tab. 1

Completed (born before 1912) and incomplete (born after 1911) family size of ever-married women (married once and before 45 years of age) recorded in the 1961 Italian census

Birth cohort Italy North-West North-East Centre South Islands Completed fertility < 1887 4.38 3.60 4.89 3.97 4.96 4.65 1887-1891 4.17 3.33 4.48 3.73 5.04 4.63 1892-1896 3.82 3.00 3.95 3.39 4.81 4.48 1897-1901 3.73 2.92 3.74 3.31 4.87 4.42 1902-1906 3.58 2.78 3.47 3.14 4.81 4.39 1907-1911 3.30 2.55 3.06 2.86 4.51 4.18 Incomplete fertility 1912-1916 3.01 2.32 2.72 2.60 4.07 3.93 1917-1921 2.72 2.08 2.39 2.32 3.69 3.66 1922-1926 2.40 1.87 2.11 2.07 3.21 3.18 1927-1931 1.98 1.56 1.76 1.75 2.56 2.59 1932-1936 1.42 1.11 1.27 1.28 1.78 1.89 1937-1941 0.89 0.67 0.82 0.79 1.08 1.20 1942+ 0.57 0.47 0.62 0.50 0.57 0.66 Total 2.64 2.08 2.50 2.32 3.42 3.29

Completed (born before 1912) and incomplete (born after 1911) family size of ever-married women (married once and before 45 years of age) recorded in the 1961 Italian census

Note: Regions included in the various districts. North-West: Piedmont, Lombardy, Val d’Aosta, and Liguria; North-East: Veneto, Friuli Venezia-Giulia, Trentino Alto-Adige, and Emilia Romagna; Centre: Tuscany, Umbria, Marche, and Lazio; South: Abruzzi e Molise, Campania, Apulia, Basilicata, and Calabria; Islands: Sicily and Sardinia.

14The North-South differential starts slowly to narrow with the generations with incomplete fertility (born after 1911), still standing at around 1.6 for the 1937-41 birth cohorts. This is clearly appreciable in figure 2, where the national, the minimum and the maximum regional values for each birth cohort are plotted. There is a clear tendency towards an overall reduction in the number of children ever born for generations with completed fertility. But there is also a general reduction of the difference between the maximum and the minimum regional value of fertility, which, after having peaked for the 1912-16 birth cohorts, rapidly narrows in the following generations. At the regional level, Liguria (and, to some extent, Piedmont and Tuscany) shows the lowest figures of completed family size over the generations, proving to be the regions forerunner of fertility decline in Italy, while Basilicata and Sardinia are the regions presenting the highest values. On the other hand, Friuli Venezia-Giulia and Emilia Romagna show the fastest pace in reduction for the birth cohorts with completed fertility (over 40% fewer children for the 1907-11 birth cohorts compared to the birth cohorts before 1887), and Sardinia emerges, instead, as the Italian region with the slowest demographic transition, [8] with a still growing completed family size.

Fig. 2

Completed (born before 1912) and incomplete (born after 1911) family size of ever-married women (married once and before 45 years of age) recorded at the census of 1961. National, minimum and maximum regional values

Completed (born before 1912) and incomplete (born after 1911) family size of ever-married women (married once and before 45 years of age) recorded at the census of 1961. National, minimum and maximum regional values

The populations studied

15The populations investigated belonged to different Italian regions, and they were selected with the aim of representing different demographic contexts (see tab. 1 and fig. 3) and socioeconomic structures. Turriaco is a little Northeastern community, situated on the plain of the Isonzo River, only 15 km from the former Yugoslavian, now Slovenian, border. It became part of Italy only after the First World War. Turriaco had 2,265 inhabitants in 1961 most of whom were involved in shipbuilding. This economic and occupational pattern is confirmed by the residential pattern. The vast majority of the population (97%) lived in the village with only a handful living out in the countryside. In our view, Turriaco should represent the rapid fertility decrease found in North-Eastern Italy, not least because of the non-agricultural nature of the community.

Fig. 3

Geographic localization of the ten communities studied

Geographic localization of the ten communities studied

16Lauco is a mountain municipality encompassing a group of villages situated in the Eastern Alps, not far from the Austrian border, at an altitude of about 800 m. It was characterized, as with many other communities in this area, by the seasonal and temporary emigration of men, especially toward France, which was necessary for supplementing the meager income deriving from a local economy based largely on subsistence agriculture. Migrants were, in large part, involved in construction activities. This poor economy explains the steady decline in population size after the First World War, which brought the inhabitants from 3,154 in the census of 1921 to 2,127 in the census of 1961.

17Novellara and Casalguidi are two populations located, respectively, in the heart of the Po plain, just under 20 kilometers from Reggio Emilia, and in the Tuscan municipality of Serravalle Pistoiese, close to the city of Pistoia. Novellara, with about 10,000 inhabitants on average, and Casalguidi, with some 4,800, have the typical pattern of rural sharecropping communities, whose economy underwent a dramatic industrial transformation in the twentieth century. However, the two populations represent two different demographic contexts: Casalguidi is located in Tuscany, one of the first Italian regions to experience fertility decline, and Novellara is situated in Emilia Romagna, the region with the steepest decline in fertility from the end of the nineteenth century. The socioeconomic change caused by industrialization, a real economic revolution in the plains of Northern and Central Italy, caused both a decline in the population living in the countryside [9] and a drop in agricultural employment [10]. Indeed, by 1961 agriculture no longer formed, especially in Emilia and Tuscany, the backbone of the economy. In particular, local peasant populations associated with the sharecropping system were disappearing (Becattini, 1975). Based on a labor-intensive rural economy, the sharecropping family relied on the domestic working force to compete for the best contractual conditions and the best farms. In turn, the need for a large domestic working force meant higher fertility, a patrilocal form of living arrangement after marriage and a joint family system. Sharecropping was formally abrogated, after years of decline, only a few years after the census of 1961: the consequences for local fertility levels can be imagined.

18As for Southern Italy, we have considered five municipalities in Sardinia – the latecomer in Italian fertility decline – and one from Campania as a comparison. The communities from Sardinia represent the various socioeconomic contexts of the island: Alghero is a town of around 27,000 inhabitants in 1961 situated on the northwestern coast of Sardinia; Porto Torres is an important port city of around 11,200 inhabitants in 1961 on the northeastern coast; Macomer is a large municipality of Central Sardinia (about 8,000 inhabitants in 1961) situated at a crossroads of rail and road routes; and, finally, Austis and Seulo are two smaller municipalities (respectively 1,482 and 1,772 inhabitants in 1961) in the Sardinian interior with agriculture economies.

19Alghero, Porto Torres, and Macomer represent the different facets of the urban society of Sardinia. While Alghero presented many facets – the socioeconomic structure of the city centre, the sea-related activities typical of the harbor, and the rural economy of the surroundings countryside – Macomer relied prevalently on industry and commerce (Brigaglia, 2008). To sum up, both populations present, in the census of 1961, a variegated socioeconomic structure. They had farmers and traders, skilled and unskilled workers, as well as people involved in transport activities, associated with the sea in Alghero and railroad companies in Macomer. These particularly attracted skilled workers, technicians, and engineers coming from Sardinia but also from mainland Italy. Another important institution for the occupational structure of Macomer was its prison, which was one of the most important in Sardinia. As for Porto Torres, its economy and social structure was strictly associated with the harbor. Not only did 20% of household heads work there, but the harbor favored the presence of many activities associated with trade. Less important than in Alghero were fishery and agriculture. On the other hand, Austis and Seulo should represent the typical rural society of Sardinia, secluded and involved in traditional rural activities (Mattone, Sanna, 1994). As they had a similar size and socioeconomic structure, they will be treated as a single population in the analysis. This selection of five communities makes it possible to investigate whether Sardinia was a unique “immobile” demographic system or rather a variegated population with different micro-demographic systems.

20The final community is Massa Lubrense, a municipality overlooking the Gulf of Naples with 9,158 inhabitants in the 1961 census living in about twenty scattered villages. The economy was mostly based on agriculture and commerce, but sea-related activities were present as well. The existence of specialized cultivation in the area was a factor limiting emigration flows in the early twentieth century (Filangieri, 1910).

21It is worth stressing that we are describing recent changes, which should be kept in mind when discussing fertility trends and changes. Most women interviewed for the 1961 census came from generations who had grown up in a world with very different social norms and customs, where agriculture was still central in the lives of most individuals.

The descriptive analysis of fertility in the nine populations

22Before describing the fertility patterns of the ten communities, it is important to stress that the analyses presented hereunder will take into consideration ever-married women with both completed and incomplete fertility. [11] In accordance with how ISTAT devised its tables, stillborns have been excluded from the analysis, and only marital reproductive history of first marriages has been taken into account.

23Figure 4 shows the trend of general fertility over time. [12] This cross-sectional measure has been computed by retrospectively projecting the reproductive history of ever-married women in the 1961 census. It is obviously only a rough measure of fertility, as it is potentially biased by different mortality patterns among birth cohorts, and, most of all, it determines a different age structure by period, the younger the further we go back in time. However, this index has been used in publications of the “Istituto Centrale di Statistica” about the vital statistics of the interwar period at the provincial level, whose values are consistent with those here presented for specific municipalities, especially the largest ones. Thus, despite the intrinsic limits of the index for the “real” levels of fertility, the plot here below allows, in any case, for a sense of trends in the ten populations studied.

Fig. 4

General marital fertility rate in the nine populations studied

General marital fertility rate in the nine populations studied

24What emerges clearly is how the evolution of fertility varies dramatically between northern and central populations, on the one hand, and southern ones, on the other. The southern populations show a slower pace of fertility reduction, especially if compared to the northern-central communities, where the drop in fertility is much more rapid. Some heterogeneity is present also within the same region, as in Sardinia, where the urban population of Macomer, and even more those of Alghero and Porto Torres have more dramatic fertility decreases than the rural populations of Austis and Seulo. Sardinia is also the region where the effects of WWI on fertility were more limited, especially if compared with the dramatic drops in the populations more directly involved in the conflict.

25A more robust and correct measure of fertility, which takes into consideration both completed and incomplete reproductive histories, might be the number of children ever born within a given marriage duration, say, ten years. Of course, ten years is often only a part of a married life, but it is a period long enough to highlight possible changes in the fertility levels and reproductive pattern, especially in a transitional phase. At the same time it is also not so long to curb substantially the number of married women involved in the analysis. [13] The descriptive statistics shown below are, in any case, based on first marriages celebrated in 1951 or before, which lasted at least ten years and whose bride’s age at first marriage was younger than 40 years. Table 2 shows the number of children per married woman by marriage cohort and population. The results are consistent with those presented in table 1 and point once again to a North-South dichotomy. While the cohorts married before 1916 present similar fertility values across Italy, the pace of fertility decline is definitely different between North and South Italy. Northern communities already show figures around the replacement level in the 1916-45 marriage cohorts, with numbers dropping clearly below that threshold after WWII: this shows the North again as the pioneer of Italian fertility decline. Sardinian communities offer, conversely, a slower pace of fertility reduction, with fertility values generally over three children per married woman in the 1946-52 marriage cohorts. Macomer is an exception, with fertility levels that are consistently lower than the rest of region. It is only after WWII that Sardinia’s demographic delay becomes more evident, since in that period even the other southern community of Massa Lubrense shows a consistent drop in fertility: Massa passes from 3.5 to 2.5 children per married woman.

Tab. 2

Number of children per married woman in the first ten years of marriage by marriage cohort and community

Municipality <1916 1916-45 1946-52 Women Children/Woman Women Children/Woman Women Children/Woman Lauco 73 4.0 235 2.5 23 2.0 Turriaco 78 3.7 353 2.2 79 1.6 Novellara 311 3.8 1392 2.3 477 1.6 Casalguidi 141 3.2 639 2.3 204 1.8 Massa Lubrense 193 4.0 789 3.5 259 2.5 Alghero 300 3.8 1488 3.5 353 3.0 Porto Torres 218 3.7 902 3.2 276 3.3 Austis-Seulo 35 3.4 172 3.4 57 3.7 Macomer 124 2.9 642 3.3 207 2.8 Total 1473 3.6 6612 3.0 1935 2.5

Number of children per married woman in the first ten years of marriage by marriage cohort and community

Note: Only first marriages with a duration of at least ten years and age at marriage under 40 years.

26The evolution of fertility levels shown in the previous table goes hand in hand with variations in the mean age at first marriage of women (tab. 3). In all the populations, age at first marriage increases, sometimes even remarkably, between the first and the second marriage cohorts considered. In the 1946-52 cohort, the increases with respect to the 1916-45 cohort are smaller, and in two cases (Lauco and Turriaco) it even decreases. The higher age at first marriage of the women living in Macomer and inner Sardinia could be one of the demographic mechanisms explaining lower marital fertility, especially as far as earlier marriage cohorts are concerned.

Tab. 3

Mean age at first marriage of women by marriage cohort and community

Municipality <1916 1916-45 1946-52 Women Lauco 23.8 25.6 25.1 331 Turriaco 22.8 24.2 23.8 510 Novellara 21.8 23.6 24.2 2180 Casalguidi 21.6 22.8 23.8 984 Massa Lubrense 23.9 25.0 25.9 1241 Alghero 21.4 23.1 24.4 2168 Porto Torres 21.1 22.7 22.8 1396 Austis-Seulo 22.0 25.1 26.2 267 Macomer 22.8 24.2 24.4 994

Mean age at first marriage of women by marriage cohort and community

Note: Only first marriages with a duration of at least ten years and age at marriage under 40 years.

27The analysis of mean age at first marriage by education attainment is also interesting. One of the demographic mechanisms put forward to explain the lower fertility of educated women is their higher age at first marriage. This is precisely what we found in our populations, where non-educated and primary-school leavers had a mean age at first marriage, all cohorts considered together, of about 23.5 years, whilst highly-educated women married rather later, on average 25.3 years.

Demographic transition at the individual level: Some determinants of fertility across marriage cohorts

28It has been widely demonstrated – not least in the descriptive analyses carried out on the 1961 census (ISTAT, 1974; Livi Bacci, 1977) – that many elements play a role in shaping the reproductive pattern of couples. However, univariate analyses are not fit for highlighting the complex role of the many factors (biological, socioeconomic, and cultural) that are potentially determinant in modifying and influencing human fertility.

29Potential predictors of fertility of ever-married women were sought using Poisson regression, with an exposure time of ten years (of marriage) for each woman. Absence of over-dispersion was checked for in each of the estimated models. [14] Coefficients were expressed in relative terms through incidence-rate ratios (IRR). All statistical computations were done using the STATA statistical package.

30The female population studied here consists of women married once before 40 years of age for a marriage duration of at least ten years. The focus is, therefore, on the number of children in the first ten years of marriage. The explanatory variables introduced in the models are addressed to capture the effects of women’s individual histories as well as their socioeconomic and cultural background. The introduction of marriage cohorts is aimed at highlighting a possible decline in the number of children across cohorts, as the above descriptive analyses seem to suggest, once controlled for age at first marriage. Birthplace and level of education focus on aspects of women’s cultural background, which have previously received much attention as key elements in fertility control. Birthplace is provided to capture the effects of immigration on marital fertility, which might involve women with different social and cultural values as well as different educational levels. The immigration of married women to the various municipalities here considered was for the most part a product of intraregional movements. To highlight possible differentials in marital fertility connected to the displacement of women, we were, therefore, forced to adopt a smaller administrative subdivision, namely the provincial district.

31For education, we used three categories: No education (reference category), primary school, secondary school and degree. This categorization was necessary because of the very low level of schooling of Italian women in that period. In Italy, compulsory education was first introduced in 1877, when five-year-old children were obliged to attend primary school up until eight. This obligation was then extended to twelve years of age in 1911 and to 14 in 1923. The law penalized parents who failed to obey these laws. However, the vast majority of children, especially from the lower classes, left school after a very few years. This ineffective application of the law, especially on the female side, was the reason for the very high levels of women without any education in our case-studies. Non-educated women represented, in fact, the majority of those married before 1916 and dropped to 17% only with the 1946-52 marriage cohorts, though female illiteracy still scored over 30% in Southern communities but Macomer. As already noted, Macomer was a special case. There were many employees and technicians there with high qualifications, working for the railroad company and the prison. Thus, the proportion of non-educated women was definitely lower compared to the other Sardinian populations, while the proportion of secondary school leavers was higher, over 14% in the 1946-52 marriage cohorts.

32Obviously, in that period education level was strongly associated with socioeconomic status (SES). This is why we needed to control for SES. Actually, SES may affect fertility through many of the same intermediate variables influenced by education: openness to new ideas and new behavior; independence from religious principles; age at marriage; infant mortality; impact of direct and indirect costs of children; and access to techniques of fertility control. Information on socioeconomic status from the census returns concern house characteristics and mother’s occupation. Unfortunately, they suffer from the well-known problems associated with the use of census information for retrospective analyses. Above all, there is the problem of assuming that the SES recorded in 1961 was also experienced through the first ten years of marriage. In this paper, house characteristics were used as proxies of socioeconomic status, such as house ownership, total number of rooms and an indicator of house quality. Homeownership was common in the northern communities of Lauco and Turriaco (over 73%), in the Sardinian municipalities of Austis and Seulo (91%) as well as in Macomer (61%). In the towns of Alghero and Porto Torres, houses tended to be rented, whereas in the populations of Novellara and Casalguidi many still had sharecropping contracts with houses given by landlords (about 30%). Lastly, the “house quality” index was constructed using the presence/absence of specific house facilities (water, toilet, bathroom, electricity, gas, and heating), based on the idea that the higher the number of facilities, the higher the quality of the house and the socioeconomic status of the family. Poor houses were considered to have no more than two house facilities, rich houses at least five facilities, whilst middle-level houses three or four. The categorization into only three groups makes the transition from one category to the other more difficult. This thereby allows us to relax slightly the assumption that women had had the same housing conditions in the first ten years of their marriage as in the census of 1961. However, if we had used information on women’s occupations, the problem would have been more or less the same. Even this piece of information does not have a time-dependent nature, and it also suffers from problems of indeterminacy, given that most married women are indicated as housewives and their husbands as being retired. A last variable included in the models concerns where the municipalities were – in Central Italy, in the North (reference category), or in the South – to control for the different stage of demographic transition in such areas. One basic model was run (tab. 4), but further models with different interactions with marriage cohorts were estimated as well. [15] This was done with the hope of investigating the possible changes over time in some important determinants of fertility. Thus, the first interaction tested will be between marriage cohorts and populations, the second between marriage cohorts and educational attainment, and the third between marriage cohorts and SES (quality of the house in 1961). The statistical significance of interactions was tested by Likelihood-Ratio Test. [16]

Tab. 4

Poisson regression of the number of children for ten-year marriages. Incidence-rate ratios^a

Variables Freq. IRR Marriage cohort (ref. <1916) 14.7 1.000 1916-45 66.0 0.854 1946-51 19.3 0.731 Age at first marriage 23.5 0.985 Birth place (ref. Province district) 82.5 1.000 Elsewhere 17.5 0.976 Education attainment (ref. None) 34.7 1.000 Primary 61.8 0.917 Lower super., Super., University 3.4 0.743 House possession (ref. Own) 44.6 1.000 Tenant 35.5 1.076 Other forms (sharecropper, etc.) 15.4 1.020 Unknown 4.5 0.975 House quality (ref. Low) 12.5 1.000 Medium 70.4 0.966 High 17.1 0.945 Geographic localization (ref. North) 8.4 1.000 Centre 31.6 0.919 Massa Lubrense 12.4 1.315 Outer Sardinia (Port towns) 35.3 1.276 Inner Sardinia 12.3 1.275 Women 10,017 Log-likelihood 18,341.8 Goodness-of-fit chi2 (p-value) 9,217.8 (0.999)

Poisson regression of the number of children for ten-year marriages. Incidence-rate ratios^a

^a In bold, p<0.05.

33The results shown in table 4 prove once again the progressive decline of fertility across marriage cohorts. The drop already appears statistically significant for women married in the interwar period, 15% fewer children within ten years of marriage when compared to women married before 1916. However, the contrast is more remarkable for women married after WWII, who had 27% fewer children. The demographic delay of Southern Italy with respect to the populations situated in the rest of the country was also expected. The Sardinian communities and Massa Lubrense had higher fertility levels (around +30%) in the first ten years of marriage compared to Friuli in the North. The populations of Central Italy (Novellara and Casalguidi) appear a step forward in the control of fertility, showing 8% fewer children than northern communities.

34Figure 5, which plots the IRRs of the interaction between marriage cohorts and populations (LR chi2 = 261.1, p-value<0.001), provides a clear description of the marked drop in marital fertility across marriage cohorts in the populations of Friuli and Central Italy. This contrasts with the permanence of a demographic system of ancient regime (at least in marital fertility levels) in the South, meaning that there are larger reproductive differentials across marriage cohorts. Interestingly, the communities of inner Sardinia seem to have had a low fertility level for women married before 1916, levels that then increased across marriage cohorts. Along with higher age at first marriage (see section 4), this result is the consequence of a large proportion of childless women in those earlier cohorts (about 10%).

Fig. 5

Interaction Marriage cohorts * Geographic localization. Incidence-rate ratios

Interaction Marriage cohorts * Geographic localization. Incidence-rate ratios

Fig. 6

Interaction Marriage cohorts * House quality indicator in 1961. Incidence-rate ratios

Interaction Marriage cohorts * House quality indicator in 1961. Incidence-rate ratios

35Age at marriage presents an overall and significant negative relationship with offspring size in the first ten years of marriage, which might be the premise of a negative effect also on completed family size. Education and SES both appear as key determinants in shaping marital fertility in the first ten years of marriage. Looking at the significant results for homeownership and house quality, it emerges, as was to be expected, that wealthy women had fewer children than poor ones. In particular, women living in well-furnished houses had 5% fewer children than women living in low-quality houses, whilst tenants had 7% more children than homeowners. Homeownership was intended as a proxy for household wealth and wellbeing. However, as it happened, homeownership took on different meanings in the various contexts: in Friuli, for instance, almost everyone had a house but general living conditions of the inhabitants were harsh. As a result the interaction with marriage cohorts was estimated only for the house quality indicator (LR chi2 = 10.2, p-value = 0.037). Figure 6 allows us to appreciate that this SES indicator was associated with a significant fertility differential only among the earlier cohorts, while its effects on marital fertility tended to fall off over time.

36Education attainment unequivocally reaffirms its decisive role in the decline of fertility during the transition. Overall, a clear fertility gradient by education level emerges, with no educated women experiencing the highest levels of marital fertility. Women with primary education had 8% fewer offspring compared to uneducated women, whilst secondary school leavers had the lowest level, with 25% fewer children after ten years of marriage. More insight into the effects of education on fertility can be gained by introducing specific interactions in the model, namely with marriage cohorts and the geographical localization of the populations under examination. The interaction between marriage cohorts and education attainment (LR chi2 = 39.3, p-value<0.001) provides a better understanding of the relationship between education attainment and marital fertility across marriage cohorts and over time (fig. 7, left plot). It is evident that women with secondary education are forerunners in fertility control, presenting, already in the pre 1916 cohorts of married women, a definitely and significantly smaller number of offspring (-23%) compared to non-educated women. This gap in marital fertility among education levels seems then to narrow across marriage cohorts, especially between primary school leavers and highly-educated women. This does not hold true for non-educated women, whose pace of reduction appears slower and whose fertility gap appears increasing over time. This seems to suggest that highly-educated women were the main actors in fertility decline in the first phases of fertility transition, but that primary school leavers were, above all, responsible for the drop in fertility across the marriage cohorts here considered, that is in the central stage of the fertility transition. Non-educated women were conversely latecomers, though showing a progressive, but slower drop in marital fertility in the first ten years of marriage.

Fig. 7

Interactions Marriage cohorts * Education Level (left) and Geographic localization * Education Level (right). Incidence-rate ratios

Interactions Marriage cohorts * Education Level (left) and Geographic localization * Education Level (right). Incidence-rate ratios

37Finally, the interaction between geographical localization and education (fig. 7, right plot) allows us to highlight the different pattern in Northern-Central and Southern Italy. The northern-central populations present clear gradients of marital fertility by woman’s education (the higher the education status, the lower fertility tended to be). In Southern Italy there were, meanwhile, marked differences in fertility only for women with secondary education and only modest variations between non-educated women and primary leavers. This could be a sign of a delay in the diffusion of education-related behavior from the upper to the lowest social strata in the South and in Sardinia in particular.

Conclusions

38In the present paper, we expanded an already large database containing information on the reproductive histories of ever-married women contained in the 1961 Italian general census. The addition of six more municipalities allowed for a more representative picture of the different processes of fertility decline across Italy: heterogeneity that concerned not only the various regions, but also populations within the same region. Sardinia, the Italian latecomer in the process of demographic transition, actually emerged as a composite region. Coastal communities had high levels of marital fertility until the 1946-51 marriage cohorts, while interior communities, such as Macomer, Austis and Seulo, had lower fertility, whose causes, other than the higher education levels of local women, need to be better investigated.

39This large representativeness, coupled with the inclusion in the analysis of incomplete-fertility women (those married for at least ten years), allowed for a clearer idea of the role of education and SES in affecting marital fertility during the demographic transition in Italy. More education and better living conditions appeared once again to be two key determinants in fertility decline, with effects, however, that differed across marriage cohorts. But if the effects of SES, here proxied by a house quality indicator, came up in the earlier phases of demographic transition, women’s education showed its effects across all marriage cohorts covered here. This result is consistent with the knowledge acquired at the aggregate level: birth control began to spread into the lower social classes only after WWI. This started in the cities and the places of first industrialization, as primary education became typical and the economic situation began to improve (Livi Bacci, 1977). According to our multilevel models, fertility control started in the highly-educated category, spread in the interwar period among primary school leavers, then only marginally touched the non-educated group after WWII. Further research would need to analyze the role of the diffusion of education in light of the strong territorial differences in the organization of the school system. These go far beyond the two classic binaries: North vs South and countryside vs town (Zamagni, 1997). Furthermore, the diffusion of fertility control started when the Church and the Fascist State pushed pro-natalist measures. We cannot exclude that these policies had some effect on the individuals’ choices, for example on the basis of the personal compliance with the principles of regime propaganda. However, even if there was personal compliance in some quarters, the communities considered here are sufficiently differentiated to allow us to affirm that fascist policies do not score better on a micro-analytical viewpoint in relation to studies conducted at the aggregate level (Breschi et al., 2014).

40The reduction in fertility differentials as education spread through the female population, especially among highly-educated women and primary school leavers, is consistent with the innovation/diffusion hypothesis of fertility decline (Dribe et al., 2014). Women marrying older was one of the demographic mechanisms sustaining such an effect. But more in-depth studies on the intermediate variables linking female education to fertility, whether through a selection on marriage, a connection with infant and child mortality drop, or to a more general independence and autonomy in every-day decisions, are needed for a clearer idea of the intimate mechanisms of fertility decline in Italy. In the end, this paper highlights the importance of the 1961 census in covering a vacuum of knowledge in Italian demographic history and, on the other, in providing more evidence and further in-depth micro-analytic studies in the theoretical discussions about how socioeconomic status affected fertility during the transition. There are, it is true, enormous difficulties in the recovery of the original documentation of the 1961 census and the intrinsic limits of retrospective analyses (here emphasized by the strong assumptions concerning the SES indicator). However, it is hoped that other researchers may use this data set because it is certainly an important and unique source for reconstructing the initial process of diffusion of fertility control in Southern Europe, today the demographic-cultural area of the “lowest-low fertility”.