1A general finding in the empirical fertility literature is that wealth and fertility are negatively correlated. [1] This fact has been proved both at micro and macro level. During industrial revolution, most countries have experienced growing per capita income and declining fertility simultaneously, at least over the last two centuries of the Second Millennium. A similar association has been also observed, even at the individual level, in the developing countries during their more recent demographic transition. This is a further confirmation of the role of nobles, wealthy and educated people, the bourgeoisie, and town dwellers as forerunners in undertaking behaviours aimed at limiting births. The situation was not different in the period of fertility decline accompanying the Italian demographic transition. In his book A History of Italian Fertility during the Last Two Centuries (1977), Massimo Livi Bacci singled out the urban and wealthy aristocracy (along with the Jewish) as the forerunning group of fertility decline in Italy, which started roughly in the last decades of the xix^th century (1978; 1986). Other works on urban differential fertility (Schiaffino, 1979; Breschi, Derosas and Rettaroli, 2003) confirmed that in the cities upper social groups did show a lower fertility than poorer social classes already in early nineteenth century.

2On the other hand, the relationship between socioeconomic status and fertility is more ambiguous and less clear for the pre-transitional phase. If children were regarded as a normal good, then one would expect richer people to have more children than poorer people. This was precisely the idea of Malthus when he argued that economic growth would have led people to reproduce at a faster pace, and hence food supply per capita would have eventually declined. However, while plausible theoretically, it is hard to find empirical support for such an argument (Jones and Tertilt, 2006).

3If differential fertility is a difficult topic to investigate for pre-transitional populations, for Italy the situation is even more complicated. The main reason lies in the lacking of reliable documentation reporting information on the socioeconomic status of families and individuals. Studies on differential fertility are therefore more numerous from the Italian political unification (1861) onwards, when the first Office of Statistics at the national level was established. Some works on urban differential fertility (Schiaffino, 1979; Breschi, Derosas and Rettaroli, 2003) still confirmed lower fertility among the upper social groups than in the poorer strata of the Italian urban society already in early nineteenth century. Unfortunately, these considerations concern only urban populations, while very few is known for pre-transitional rural communities. The few studies dealing with Italian rural populations address the hypothesis that fertility was higher in the wealthier groups than it was in the poorest ones. Thus, in the first half of the xix^th century opposite fertility patterns by social group characterized cities and the countryside (Livi Bacci and Breschi, 1991). Breschi (1984; 1985) found, for two early-nineteenth century rural populations of the province of Pistoia, a (mild) positive relationship between marital fertility and socioprofessional group, with landowners showing higher fertility rates than and day-labourers. This pattern was found to exist also in the second half of the xviii^th century in the communities living in the countryside of Prato, a small city in Central Tuscany (Della Pina, 1993). Sala (1976) found the same relationship between social group and fertility for a coeval Northern rural community with a different social structure.

4A positive relationship between income and fertility was argued to exist also in other rural European populations early on in their development process, as proved by evidences coming from some rural populations in France (Ganiage, 1988; Weir, 1995; Fauve-Chamoux, 1997). However, in both French and Italian communities this relationship was never found to be particularly marked. [2] In rural Germany, Knodel’s study (1988) on some peasant communities did not reveal the existence of any differential fertility: “There appears to be little consistency among different villages with respect to occupational differential in marital fertility. Perhaps the most striking feature for couples married between 1750 and 1849 is the general lack of any substantial difference between occupational grouping.” (Knodel, 1988, 296). Once again, more pronounced differences emerged only at the beginning of fertility decline when the farmers started to control their reproductivity earlier than proletarians did. Similar results were observed also for three adjacent Catholic villages of the Rhine valley between 1650-1900 (Benz, 1999). Even for England, the rich nominative documentation studied by the Cambridge Group (Wrigley et al., 1997) did not provide any support to the hypothesis of differential fertility by socioeconomic status. Nor significant differences by economic structure of the parish, nor very marked and stable differentials by occupation were actually pointed out.

5A completely different situation has been recently documented for the United States (Jones and Tertilt, 2006). This study, based on 1% sample of the U.S. Census data, describes the history of the relationship between fertility and key social-economic indicators at the individual level for women born between 1826 and 1960. The most intriguing result is that already in the mid-1800’s, that is well before the beginning of demographic transition, wealth and fertility showed a negative relationship.

6Thus, if empirical evidences are not always consistent and somehow contradictory about the relationship between fertility and wealth in the pre-transitional rural populations, providing an interpretative framework is even harder because the number of living children born to women was influenced by an astonishingly wide range of factors, from physiological to socioeconomic, from behavioural to cultural. In addition, fertility is much affected by demographic factors, among which infant mortality is one of the most important. It could influence the relationship between fertility and wealth in two different and opposite ways. On the one hand, a negative relationship might result by the lower infant mortality that usually characterizes mothers living in high-income families. On the other hand, the level of infant mortality was strongly affected by different breastfeeding practices, perhaps the most frequently cited explanation for differential fertility patterns in pre-transitional rural populations. The custom of wealthy mothers to put their children out at nurse or to wean babies very early because so busy with the farm management (Bengtsson, 1999) caused both an increase in the risk of infant death and an anticipated exposure to the risk of new a pregnancy, thereby increasing fertility. In the same time, the fertility of poor nurses decreased due to prolonged periods of breastfeeding. The fertility of indigent women could be also negatively affected by heavy workloads that in turn could cause higher risks of sub-fecundity, infertility, and spontaneous abortions. At the end, many factors and elements interplay one another in ways that can affect even in opposite directions the relationship between fertility and wealth.

7Some years ago, just following upon the results of studies on urban populations, Antoinette Fauve-Chamoux wondered whether it was possible to identify a demography of wealthy and a demography of poor people also in rural societies of ancient regime (1993). Due to the few data at our disposal for historical populations, the question is not easily answered, yet. Those who tried were compelled to introduce some simplifications according to the quality of information available. Normally, researchers use data on occupations and professions to infer the socioeconomic status of individuals and therefore to discriminate between rich and poor groups. Nevertheless, this choice clashes with theoretical problems and lack of efficiency and accuracy of reconstructed data. Efficiency might be somehow compromised because parish registers, which represent the fundamental source for demographic data on pre-transitional Europe, report occupational information only in a random and unsystematic way. It is therefore highly improbable to retrieve continuous information on the professions of individuals. A second important problem that could limit the accuracy of those reconstructions is the high level of in-determination that frequently characterizes the recorded professions. For example, the generic indication of “farmer” is easily attributed to the very large part of the household heads living in a rural setting, without mentioning whether one was a sharecropper, a tenant or a day-labourer. Civil censuses scattered across Italy supply more precise and reliable data, but they are so temporally and geographically sporadic as to limit by far the conclusions that might be drawn from those sources.

8From a theoretical point of view, the use of occupational data does not always return a realistic picture of the differentials by economic status in the population. Especially in those areas of Italy characterized by a wide presence of large and complex households—i.e. sharecropping communities—the well-being of a single couple was strictly linked to the overall household economic status rather than to the individual occupation of the husband of that married couple. The type of household in which the couple lived, both from a co-residential and socioeconomic point of view, assumes therefore a strategic importance in the definition of their wellbeing and richness. Thus, differential fertility according to individual occupation/profession could not correspond to the actual economic stratification of the population. Second, in many communities people could have more than one occupation according to the season of the year. In some mountain areas of Northern and Central Italy (Fornasin, 2005), people were involved in agricultural labours during the warm season turning to other more commercial and artisan activities during the winter. Collection of scattered information on professions can therefore easily fail in taking into account this peculiar characteristic of specific social groups.

9We are fully aware of the fact that occupation, household structure and nuptiality patterns have been frequently proved as to be determinant factors in changing levels and rhythms of fertility. However, the opportunity to study the effects of wealth and richness net of those more strictly associated to the professional activity or occupation might be useful, for example, to have some more clear indications on the role of the nutritional state in affecting fertility and fecundability.

10What we aim at analyzing in this paper is to what extent wealth could influence fertility in a rural pre-transitional Tuscan population—Casalguidi—between 1819 and 1859, where the differentials in wellbeing are expected to be small due to the large prevalence of agricultural labourers in the population. Along with this aggregate analysis, an individual-level approach will be employed to test whether the economic status was an important determinant of fertility after controlling for other intermediate variables.

The sources

11The data used in this study comes from both religious and civil documentation. Ecclesiastic sources are annual nominative lists of inhabitants called Status Animarum, supplemented by parish registers of vital events covering the period 1819-59. The Status Animarum was a sort of local census recorded by the parson each year around the time of Easter. The Status Animarum supplies information on age, sex, marital status, and relationship to the household head of each individual living in the parish territory. This latter information is possible since these registers are organized by household. Records of vital events (such as baptisms, burials, and marriages) extracted from parish registers were then linked to the information from Stati Animarum thereby making possible to define year by year the household structure where the ego lived as well as to trace his/her movements within the community. This integration of census and vital data allows us also to situate the birth of a child in the context (individual, couple, and family) where he occurred (Manfredini, 1996; Breschi, Derosas and Manfredini, 2004).

12As for socioeconomic information on household and people, the Status Animarum does not provide continuous information on household socioeconomic status. Apart sporadic information on the profession of the head, such registers always reported name and surname of homeowners. Comparing this piece of information to the name of the household head, we were able to determine whether the household owned its home. Other data on occupation can sometimes be found and drawn from vital registers, for instance on marriage acts for spouses or birth acts for parents.

13More significant information on the socioeconomic status of households were obtained from the Tax Register (Breschi, Manfredini and Pozzi, 2004). This civil source, which covers the same period covered by religious documentation, supplies information on the name and surname of the household head, his/her occupation, hamlet of residence, level of taxation, and other details for each taxable household. The law provided that the total amount deriving from taxes, fixed by the central government, had to be divided among those heads deemed able to pay by the representatives of the community. [3] All “miserable and needy families” were therefore exempt from taxes. The same law provided the possibility of deduction in case of emigration, insolvency, and intervened widowhood. [4] Since the number of tax classes changed over the period studied, we managed to have only four levels of taxation per year. [5]

14This archival register offers a unique documentation for assessing the socioeconomic status of the families of a community and the linkage between Tax Register and Status Animarum provides a reliable picture of the socioeconomic hierarchy that fits our purposes. [6]

The population studied and its socioeconomic structure

15Casalguidi is a large village in the province of Pistoia that, during the period of our study, was administratively part of the Grand Duchy of Tuscany. Between 1819 and 1859, the population averaged around 2,400 inhabitants, showing a constant growth despite the cholera epidemic of 1854-1855 (Breschi and Manfredini, 1998).

16Like many other communities of eighteenth-century Tuscany, Casalguidi was primarily a rural society where about 75% of adult people were employed in agricultural labours, although some forms of proto-industry, such as silk weaving and embroidery, were present and growing up in the territory. Despite this apparent uniformity, the various agricultural categories were marked by considerable differences in demographic behaviour, migratory attitude, family formation systems, household structure, and, finally, socioeconomic status (Barbagli, 1984; Kertzer and Hogan, 1991; Viazzo and Albera, 1992). Especially day labourers, on the one hand, and sharecroppers, on the other hand, represented the traditional dichotomy of the Tuscan countryside. Sharecroppers married later, followed a patrilocal system of living arrangement after marriage, and lived in large and complex households. On the other hand, day labourers married earlier, had a neolocal family formation system, and lived in simple family groups, formed by only one biological nucleus (Della Pina, 1990; Barbagli, 1990; Rettaroli, 1993). The differences in the marriage pattern were the consequence of a different tie with the land (Poni, 1982; Doveri, 2000). Sharecroppers lived on the farm they cultivated for an absent landowner. The contract tied the whole family group to the landowner and provided the equal division of the crop between the sharecropper and the landowner. No member of the sharecropping household was allowed to work outside the farm. Each year the contract had to be renewed and one of the key points was the capacity of the sharecropping household to guarantee an adequate quantity of crop to the landowner (Giorgetti, 1974; Pazzagli, 1973). Thus, one of the main concerns for sharecroppers was to preserve an adequate working force within the household by adopting specific demographic behaviours, such as higher fertility, expulsion of exceeding and less productive members, and a patrilocal arrangement after marriage for men.

17Day labourers had no tie with the land since they did not live on the farm, but moved around in search of farmers offering them the opportunity to work for some time. When the demand for agricultural labours declined, day labourers could find employment in poor artisan activities, so that the two groups were so easily interchangeable that they had similar nuptiality patterns and family formation systems. Mobility was their main characteristic, along with neolocality. For day labourers, the familial labour force was not the central factor for finding or maintaining a job, and large households were not fit for their frequent movements and unsustainable for the resources of this social category. As a consequence, the members of those nuclear households, both men and women, left the native family on marriage.

18It is normally accepted that sharecroppers were in better economic conditions than day labourers were, but the following tables tell a different story. If we classify households by head’s profession and tax amount as recorded on the Tax Register, there would not seem emerging any differential in wealth between day labourers and the other farmers, and even artisans (Table 1). All these social groups show a proportion of households falling into the two lower classes of taxation between 79 and 83%, indication of an overall low economic status. [7]

Tab. 1

Households by Head’s Profession and Tax Class (%). Tax Register, 1819-59

Profession High Tax Medium Tax Low Tax No indication Total % Total N Day labourers 1.1 16.4 82.1 0.4 100.0 3 358 Sharecroppers & other farmers 1.7 14.9 83.2 0.2 100.0 6 138 Artisans & other non-agricultural activities 2.5 18.8 78.5 0.2 100.0 1 761 Nobles, landowners & middle-class 55.7 28.0 15.8 0.5 100.0 326 Total % 3.3 16.2 80.3 0.2 100.0 11 583 Total N 382 1 876 9 302 23 11 583

Households by Head’s Profession and Tax Class (%). Tax Register, 1819-59

19Conversely, the most outstanding professions and occupations result well differentiated in terms of wealth, with 83.7% of households paying the highest tax amounts. The No Indication column refers to households whose tax indication was missed or impossible to understand on the source, or that had been exempted from paying taxes, in the case of an intervened widowhood, for instance.

20Since Tax Registers do not include tax-exempt households, we tried to recover information on the profession of those household heads from the parish registers to check whether a larger part of tax-exempt households could belong or not to the group of day labourers (Table 2). The picture above described is confirmed: day labourers and farmers do not differentiate one another in terms of household wealth, with respectively only 12.5% to 12% of families living in good economic conditions and about 29% of tax-exempt households due to manifest poverty. In general, the number of indigent and needy households was very high, over 85%, denoting a situation of diffuse poverty.

Tab. 2

Households by Head’s Profession and Tax Class (%). Tax & Parish Registers

Profession High Tax Medium Tax Low Tax Tax Exempt Total % Total N Day labourers 0.9 11.6 58.6 28.9 100.0 4 723 Sharecroppers & other farmers 1.3 10.7 59.3 28.8 100.0 8 621 Artisans & other non-agricultural activities 1.9 13.6 56.9 27.6 100.0 2 432 Nobles, landowners & middle-class 47.4 23.9 13.5 15.1 100.0 384 No Profession 0.3 4.2 13.0 82.4 100.0 2 393 Total % 2.1 10.7 51.9 35.3 100.0 18 557 Total N 386 1 993 9 625 6 549 18 557

Households by Head’s Profession and Tax Class (%). Tax & Parish Registers

Giving birth in Casalguidi: some descriptive results

21Through the techniques of population reconstruction described in the previous section, 12862 person-years (married women 15-49 years) and 3662 births were observed. The mean number of children per married woman was actually 10.4, in line with the intensity of fertility found for other communities of eighteenth-century rural Tuscany (Salvini, 1997; Livi Bacci and Breschi, 1991). The figures for TMFR₂₀ and TMFR₂₅, which are parameters more indicative for populations with late marriage such as Casalguidi, are 8.6 and 6.3 children per woman, respectively.

22If TMFRs are computed on ten-year periods, there is a certain steadiness around nine children per woman, with the exception of the last decade (1850-1859), when TMFR₂₀ drops to only eight children per woman. This result is easily the consequence of the last great cholera epidemic that spread across Northern Italy in the biennium 1854-1855.

23Figure 1 shows the curve of age-specific marital fertility rates. It depicts a population not yet controlling its fertility, with age-specific fertility rates peaking at 20-24 years and showing a marked drop from 40 years onward. Confirmation of the absence of deliberate birth control comes from computation of the Coale-Trussell’s m. The resulting value of 0.132 is far from suggesting the existence of even some degree of voluntary birth control. [8] An examination of total marital fertility rates by age at first marriage helps to clarify the situation (Figure 1). [9] Women married at 25+ years show higher marital fertility than those who married younger. Women who married later may have had a stronger desire to exploit their remaining opportunities for reproduction. On the other hand, the marital fertility of those who married young may have been reduced by physiological consequences due to earlier pregnancies. Since the mean age at first marriage was around 25 years, the latter pattern represents the most common situation among the married women of Casalguidi.

Fig. 1

Age-Specific Marital Fertility Rate. All Marriages and by Age at Marriagea

Age-Specific Marital Fertility Rate. All Marriages and by Age at Marriagea

24As for the occupational groups, the two most important groups of farm workers were analyzed. The large category of farmers (sharecroppers, tenants & smallholders) showed a slightly higher marital fertility in comparison to day labourers, 8.2 against 7.7 children per woman. However, in order to have a more precise classification of professions, the category of sharecroppers & other farmers was splitted into two groups according to the propriety of the house where they lived. This allowed us to separate all those farmers that cultivated farms for absent landowners, such as sharecroppers and tenants, from smallholders and other farmers that were not hired by landowners. The results are in line with data on other rural communities, with sharecroppers displaying higher marital fertility than farmers with the house and day labourers. However, the differences are not striking as the TMFR₂₀ are 8.3, 7.8, and 7.7 children per married woman, respectively. This narrow difference is however not surprising since the wealth status among the social groups here considered was similar. It is to say that the constraint imposed by landowners in terms of balance between landholding size and household size forced sharecroppers to control an excessive growth of family size by both limiting the access to marriage of women, whose celibacy rate was around 10.5% in Casalguidi, and increasing female age at first marriage. Since in Casalguidi no substantial differences between sharecroppers & other farmers and day labourers in both age at first marriage and age at last birth were found, the slight differential in marital fertility between the two groups was due to a different birth spacing (Table 3). Farmers living on the farm had shorter intergenesic intervals than day labourers had, respectively 2.1 and 2.4 years on average.

Tab. 3

Some Features of the Fertility Pattern by Agricultural Social Group

Social Group Age at first marriage Age at last birth Protogenesic interval Intergenesic interval Sharecroppers & other farmers 25.0 39.7 1.1 2.2 Day Labourers 25.1 39.8 1.1 2.5

Some Features of the Fertility Pattern by Agricultural Social Group

25In the end, the dichotomy sharecroppers/day labourers does not solve the question of the existence of a fertility pattern of poor and a fertility pattern of rich couples: the intensity of fertility was actually not so dissimilar in the two social groups and so were their wealth standards.

26Let us turn then to fertility by family tax, that is household resources and income. Figure 2 responds quite clearly to our question. Fertility is positively associated to the household wealth level with a gradient going from the poorest and low-fertility to the richest and high-fertility households.

Fig. 2

Age-Specific Marital Fertility Rates by Family Tax Class

Age-Specific Marital Fertility Rates by Family Tax Class

27The above pattern holds at each age from 20-24 to 45-49 years, with the sole exception of the Low-Tax and Untaxed classes whose age-specific marital fertility rates overlap up to 30-34 years. In terms of TMFR₂₀, it passes from 10 children per married woman of the High-Medium Tax class to 8.5 of the Low-Tax class and, finally, to 8 of the untaxed households. This entails that the final offspring size of rich couples was about 25% higher than that of indigent ones. Even this time, in fact, differential fertility was a question of spacing rather than stopping. In table 4, we have calculated the same indicators of table 3 differentiated by taxation level also adding the infant mortality rates (IMR).

28Mean ages at marriage and at last birth did not vary significantly among the tax groups once controlled for age at marriage. As before, poor married women showed longer birth intervals than women in wealthier households, especially as far as late marriages were concerned. IMR does not help in finding an explanation. The lower infant mortality of upper classes was associated to shorter intervals, whilst the higher level of infant mortality among the poorest ones combines with longer intergenesic intervals. This is unexpected since, in a natural fertility regime, usually the higher the IMR the shorter the interval between two consecutive births, and, consequently, higher was marital fertility. Since wet-nursing was a common practice in Casalguidi, it is possible that wealthy women put sometimes some of their babies out to nurse, thereby reducing their length of breastfeeding and returning fertile sooner than expected. We found the names of more than 110 wet-nurses in the acts of infant dead, and almost each of them belonged to the two lowest tax classes. This fact was not a surprise as mercenary wet-nursing was practiced most of all to supplement the meagre household incomes of poor families. [10] This situation may in turn have played a role in lowering their marital fertility due to the delay in ovulation following a prolonged lactation even in presence of high infant mortality (Oris, Derosas and Breschi, 2004; Leridon, 1984; Corsini, 1974).

Tab. 4

Indicators of the Reproductive Period by Household Taxation Level

Taxation level Mean age at first birth Mean age at last birth Mean birth interval IMR Married 20-24 Married 25-29 Married 20-24 Married 25-29 Married 20-24 Married 25-29 High & medium tax 22.7 27.1 40.6 41.9 1.78 1.79 172.0 Low tax 22.7 27.4 39.4 41.0 1.95 1.98 195.2 Untaxed 22.8 27.2 40.0 41.6 1.97 2.11 190.6

Indicators of the Reproductive Period by Household Taxation Level

29However, other factors may be at play in affecting negatively the fertility of poor couples: mobility and malnutrition. The higher mobility of day labourers, frequently moving in search of work, could cause longer periods of husband-wife separation thereby reducing their marital fertility, a pattern that has been observed in Spain and Swiss Italy among others (Oris, 2003). [11]

30We can also hypothesize the influence of malnutrition and bad diet (Alter, Oris and Neven, 2007). Despite the impossibility to collect reliable household-level data on food consumption and diet of the families living in Casalguidi, some preliminary and indirect indications can be drawn from the analysis of infertile couples. If we assume infertility to be a consequence of malnutrition and depletion, we could suppose infertile couples be more frequent among the poorest sectors of the population. [12] For simplicity, we regarded as infertile those couples at their first marriage, under observation for at least 10 years from marriage (more or less two-third of the reproductive period) and with no children. The results seem going in the expected direction: about 7% of couples living in untaxed households had no children in their first 10-year period of marriage, against proportions of around 1% among the other two tax classes. [13]

Individual-level analysis of fertility: the role of wealth and economic factors

31As shown in univariate analyses, many factors contributed to make complex the social structure of the peasant community of Casalguidi, namely profession, wealth, house property, household structure, mobility, family formation system, and so on. To separate their respective roles and impacts on the woman’s risk of bearing a child a multivariate approach was used. The role of wealth in affecting fertility could be therefore more clearly interpreted—i.e. claiming the effects of malnutrition—since it will be possible to disentangle the effects actually connected to specificities of each single occupation and its tie with the land from those deriving from the wealth status. Here, we used Event History Analysis as it is one of the most powerful statistical tools to deal with individual-level data. The dependent variable is therefore a dichotomous variable indicating if a given married woman had or not a child in year t. Explanatory variables describe individual, household, economic and context characteristics of that woman at the beginning of the same year. Currently married women between 15 and 49 years whose marital reproductive history has been reconstructed from its start form the population at risk.

32Besides the Class of Taxation and the Household Head’s Profession, we include other three important factors that might play a role in affecting fertility and that could interplay with the previous two variables, namely household structure, the wheat price, and wet-nursing activity of the ego. A positive association between household complexity and marital fertility has been already detected in previous studies on Casalguidi (Breschi et al., 2000). Household structure had close ties with household wealth. In 1927, Livio Livi was one of the first scholars to support the thesis of a strict positive relationship between socioeconomic status and household size in agricultural populations of mid-nineteenth century Italy. In a rural economy where farmers owned usually only small plots whilst absent landowners owned the larger the estates and the household, the more productive the farm that the household could access to. However, size was not a value per se, since composition was determinant as well. The availability of human labour and the capacity to supply quality labour, features that can be found in a large number of working-age male members with solid farming experience was actually key in getting the best and largest farms (Doveri, 2000). This principle was so strong that small sharecropping households were likely to be evicted from the farm with consequent displacement to a smaller one or downward social mobility. This situation is clearly showed in figure 3, where we classified households by structure and tax class.

Fig. 3

Households by Tax Class and Structure

Households by Tax Class and Structure

33The proportion of households belonging to the upper tax classes decreases passing from multiple to simple households to, even much more, solitaries. Conversely, percent figures of households in the untaxed group follow the opposite direction, with almost 39% of simple households exempt from tax for manifest poverty.

34The variable Price of Wheat is an annual average based on the prices of the close city market of Pistoia and it has been lagged by one year to describe the economic situation at the moment of conception. It represents a useful indicator of economic conjuncture and short-term economic stress (Bengtsson, 2004). The purpose is to check the existence of any differential response by profession and/or wealth level to variations in prices. In a previous work on Casalguidi, we found that child mortality was much more responsive to price changes among the poorest than among the other economic categories (Breschi, Manfredini and Pozzi, 2004). Here, we want to test if the same hypothesis holds as far as fertility is concerned (Galloway, 1988; Bengtsson and Dribe, 2006).

35As mentioned above, wet-nursing was a common practice among the poorest classes, thereby potentially interfering with their fertility. Its inclusion in the models aims at evaluating the relationship between the risk of childbearing and wealth net of that important element. Finally, age and age at marriage have been included in the models to control for both the biological pattern of fertility and the timing of the beginning of the marital reproductive period. For the variable relative to the profession of the head, we separated the category of farmers into farmers living in their own house and those living in the landowner’s house.

36Three models have been estimated. The first one analyzes the effects of head’s (husband) profession on fertility once controlled for age, age at marriage, economic short-term stress (wheat price), and activity of wet-nursing. In the second and third models model we add household structure and tax class, respectively. In this way, we want to assess the impact and role of those two factors on fertility once controlled for the other variables above mentioned.

37Table 5 shows the relative risks for each of the categories considered in the analysis. The results are clear, confirming Livio Livi’s hypothesis: fertility is affected by both household structure and wealth, with profession remaining in the background. This latter variable plays a significant role in affecting the risk of childbearing only as long as household structure, first, and tax class, after, are included in the model. Thus, the significant higher risk of giving birth for sharecroppers & tenants in comparison to day labourers was actually related to the larger and more complex structure of the households in which sharecroppers lived. Some authors have seen in this higher fertility of large family groups a higher demand for children triggered by the necessity to guarantee themselves sufficient labour-force supply (Doveri, 2000).

Tab. 5

Effects of Socioeconomic Factors on the Relative Risk of Birth for Currently Married Women Age 15-49. Casalguidi 1819-59

Model 1 Model 2 Model 3 Variables Mean Relative risk Woman’s age 15-19 1.2 1.009 0.946 0.947 20-24 24.2 1.245 1.194 1.201 25-29 (ref.) 13.6 1.000 1.000 1.000 30-34 23.0 0.807 0.828 0.830 35-39 17.2 0.594 0.624 0.624 40-49 20.7 0.161 0.174 0.173 Woman’s age at marriage 15-24 23.5 0.827 0.848 0.848 25-29 (ref.) 67.8 1.000 1.000 1.000 30+ 8.6 1.352 1.361 1.380 Wet-nurse No (ref.) 79.4 1.000 1.000 1.000 Yes 20.6 1.117 1.140 1.158 Occupation of household head Day labourers (ref.) 24.8 1.000 1.000 1.000 Sharecroppers & tenants 37.3 1.132 1.090 1.099 Other farmers in their own house 13.0 1.171 1.145 1.120 Artisans / shopkeepers 14.0 1.128 1.104 1.080 Nobles & Middle-class 1.6 1.375 1.368 1.291 Unknown 9.3 1.173 1.133 1.144 Lagged wheat price (logged) 3.1 0.780 0.791 0.806 Household structure Simple (ref.) 47.8 1.000 1.000 Extended 26.1 1.122 1.087 Multiple 26.1 1.289 1.232 Taxation level High & medium 13.6 1.236 Low (ref.) 62.3 1.000 Untaxed 24.1 1.020 Model 1 Model 2 Model 3 Person years 6,148 6,148 6,148 Number of births 2,091 2,091 2,091 log-likelihood -3701.7 -3695.4 -3692.4 LR chi2 - p-value 12.71 - 0.002 6.00 - 0.049 Note: in bold p ? 0.05

Effects of Socioeconomic Factors on the Relative Risk of Birth for Currently Married Women Age 15-49. Casalguidi 1819-59

38As for wealth, measured through the tax each household had to pay, its effects on fertility were clear. Once controlled for other factors potentially influential, economic conditions did affect the risk of having a baby, conclusion that is supported, from a statistical point of view, by the significant increase in the log-likelihood once the taxation level is introduced in the model. However, the single coefficients tell us that wealth affected fertility only when economic conditions were by far over the local standards. Actually, women living in households with a very good economic situation showed a 25% significant higher risk of childbearing than day labourers did. On the other hand, no differential risk was detected between women living in low-tax households and women living in untaxed households.

39As for wet-nurses, the result is unexpected. Unlike what was argued, when controlled by the other variables, wet-nursing increases the risk of childbearing by a significant 16%. One possible hypothesis concerns a possible selection bias. In fact, we identified only those wet-nurses whose names were reported on the death certificates. Presumably, they are only a selected part of all women practicing that activity, in particular those wet-nurses who lost their nursed children very early. Thus, we are capturing fertile women who had recently delivered, otherwise they could not be employed to suckle an infant, but whose breastfeeding was interrupted by the dead of the nursed child. The anovulatory period was therefore shorter than it was expected for women practicing wet-nursing and this could partly explain the positive relationship we found between wet-nursing and fertility in this case-study.

Conclusions

40Our study on a Tuscan sharecropping community of mid-nineteenth century allows us to give preliminary answers to some open questions on differential fertility in pre-transitional rural populations.

41In a typical demographic system of ancien regime, characterized by natural fertility and absence of deliberate birth control, we find that woman living in wealthy households did have a higher fertility than those in poorer ones. This evidence, coupled with the finding of a lower infant mortality among rich households (Breschi et al., 2004), proves that in Casalguidi existed differential demographic regimes according to the household wealth, even before the transition started. This would seem confirming the hypothesis put forward by Massimo Livi Bacci, who argued that rich people were forerunners in the decline of fertility just because they were those much in need to relieve the household from the heavy family load represented by a high number of children. [14]

42Furthermore, we demonstrated that the use of profession to study differential demographic behaviours in rural populations could be sometimes misleading because it does not always reflect the real economic status of households and family groups. When controlled for household size and household wealth in a multivariate model, profession ceases in fact to be a key factor in affecting fertility.