Living Arrangements and the Elderly: An Analysis of Old-Age Mortality by Household Structure in Casalguidi, 1819–1859

Abstract

The elevated levels of protection, assistance, and care enjoyed by the elderly living in complex households has long been a key assumption of many family system theories. However, although this hypothesis has been demonstrated for contemporary contexts, quantitative evidence for past populations is particularly scarce, if not nonexistent. This article investigates the relationship between old-age mortality and living arrangements in a mid–nineteenth century Tuscan population, where the joint family system of sharecroppers coexisted alongside the nuclear system of day laborers. Our findings demonstrate that within complex households, the complexity of relationships, gender inequalities, and possible competition for care and resources among the most vulnerable household members—namely, the elderly and the young—weakens the assumption that the elderly benefitted from lower rates of old-age mortality.

Introduction

Over recent decades, the rapid aging of Western populations has drawn the attention of a number of scholars from various disciplines on the living conditions of the elderly. In particular, the increasing proportion of the elderly who live alone has prompted much research into investigating the relationship between living arrangements and old-age mortality (Grundy 2011; Iecovich et al. 2011; Li et al. 2009; Martikainen et al. 2008; Pizzetti et al. 2005). However, despite the fact that one of the key assumptions of family historians is that the quality of life of the elderly was strictly associated with the form of their living arrangements, the literature on this argument concerning historical populations is far from exhaustive. Following the first pioneer studies by the upholders of the modernization theory (Burgess 1916; Goode 1963; Le Play 1884), which support the thesis of a strong negative relationship between economic development and family complexity, elderly individuals living in large and complex households have been largely viewed as advantaged compared with those in simple and nuclear family units. It has been noted that not only could the elderly living in complex households count on a more favorable and “secure existence through co-residence with their children” (Kertzer 1995:368), being provided with assistance by a surrounding network of married children and kin, but they were also commonly attributed with high status within the family, with older men frequently exerting great internal authority and power as household head. It has therefore been held that industrialization and urbanization swept this world away through disruption of the traditional complex household system, leaving the elderly in an isolated and marginalized position, with deterioration in their living conditions. Even though the modernization theory was largely criticized by Peter Laslett and his colleagues in the Cambridge Group (Laslett 1965, 1972), the idea of a protective role played by large and complex households on the lives of elderly still persisted in the “revisionist” theory. This supports the “nuclear hardship hypothesis,” which points to the social and economic vulnerability of simple family units, especially in relation to the elderly (Laslett 1988). The assumption is that the elderly in strictly nuclear-based societies are destined to solitude and economic privation, abandoned by children who left home to find the necessary resources to get married. Laslett claimed, as a logical consequence, that an inverse relationship “exists between the size and structure of the co-residential family . . . on the one hand, and the development of an extensive public welfare system on the other” (Cavallo 1998:91). These concepts were then resumed and developed by Reher (1998), who made the relationship between the elderly and household structure more explicit. Through consideration of age of leaving home and how families manage the support of elderly members, he drew a clear-cut distinction between northern and southern European social models. Strong family systems are seen as characterized by late departure from the original family home and strong family ties, although Reher was not referring merely to the joint family system based on intergenerational complex households but to more general cultural considerations regarding kin ties and intergenerational relationships. In strong-tie families, he stated, “the care of the elderly fell almost exclusively on the family, whether this occurred through means of co-residence, the circulation of the elderly among the households of their offspring, or spatial proximity between homes” (1998:209). Reher went as far as claiming that Tuscany was one of the few European regions where the joint family system existed, functioning as a counterbalance to the underdeveloped public welfare system¹ that induced people, especially the weakest (young and elderly), to seek support within the family group and extended kin network.² More recently, Derosas and Saito (2002:3) reaffirmed such an assumption by stating that the question was “whether or not, in societies with stem or joint family organizations, the coresident domestic group could have performed the function of providing security for unfortunate individuals such as widows . . . . On the face of it, the answer seems to be yes.”³ Other studies have addressed the problem of the relationship among household structure, the elderly, and their well-being, although findings have not been consistent. Alter (1996) found that in the urban context of Verviers (Belgium), the higher the socioeconomic status, the more likely was the coresidence with children. In this context, the elderly could find some support, especially from their unmarried children, while childless elderly people faced downward social mobility. Campbell and Lee (1996) focused more directly on elderly mortality and found that in nineteenth century China, the presence of kin members did not affect the chances of survival of old-age members. Tsuya and Kurosu (2000) found, for rural Japan, a positive relationship between elderly female survival and number of kin members but no association with specific coresident kin. Evidently, the results are inconsistent, but profound differences in the context (urban vs. rural) in the sample studied (only ever-married people or the entire old-age population) and in the informative detail of the household structure may be responsible for these apparent contradictions.

The Italian situation actually proves much more complex, and the variability and co-presence of a number of family systems in the same population in any one place was the norm in Tuscany, as in the rest of Italy (Viazzo 2003). It is precisely this peculiarity of the Italian family system that allows us to analyze the effects of different living arrangements on old-age mortality, a topic that unfortunately has received little explicit attention.

In this article, we will investigate the old-age population of the mid–nineteenth century Tuscan community of Casalguidi in order to shed light on whether differences in the living arrangements determined significant differences in old-age mortality. This case study takes advantage of two key elements. First, Casalguidi, located in the heart of the Tuscan sharecropping area, was characterized by the co-presence of two contrasting family systems: the joint family system (typical of sharecroppers) and the nuclear system (typical of day laborers and poor artisans). This rather unusual situation (at least by Western European standards) with both strong and weak family systems in the same population provides a unique opportunity for investigating the degree to which differences in household structure could affect old-age mortality. Second, the documentation used enables us to determine in great detail the household composition year after year, providing great insight into the dynamics of family groups.

Household Structure and the Elderly: The Case of Rural Tuscany

Tuscany is acknowledged as a European region with a particularly long history of large and complex households.⁴ Documentation of the joint family system dates back to the fifteenth century with studies by Herlihy and Klapisch-Zuber (1978) on the Florentine Cadastre of 1427, which revealed that a large proportion of households, especially in rural contexts, were headed by senior males who continued to live with and hold authority over their married sons. Numerous studies have demonstrated that for a large part of the Tuscan population, this family formation system and family form survived for almost 500 years, up to the mid-twentieth century (Barbagli 2000; Cioni et al. 1997; Viazzo 2003). However, whereas this family system was initially characterized by an early female age at first marriage, over the last two centuries it was, conversely, associated with late marriage for females (at around 25 years of age) and high proportions never marrying for both sexes (Barbagli 2000; Della Pina 1993; Kerzter and Saller 1993).⁵ Anthropologists explain the persistence of this coresidential joint family model in cultural terms,⁶ while population scholars and historical demographers place greater emphasis on the role of economic factors, especially in relation to the Tuscan system of mezzadria (sharecropping). Sharecropping was a form of land tenure based on an equal sharing of farm produce between the landlord and tenant’s family, which contractually tied the entire household (Biagioli 1986; Francini 1984; Poni 1978). Under these premises, the sharecropping household was both a production and a consumption unit. The landowner assigned land according to the sharecropping household’s size and its labor capacity. Therefore, the sharecropping family’s well-being and the landlord’s economic return were directly dependent on the ability to maintain an unvarying workforce over time. The right balance between farm size and workforce was so essential that it was a fundamental clause in the annual renewal of the contract. Sharecropping household heads (and landlords) therefore exercised strict control over the demographic behavior of family members, and one of the most effective ways of preserving a stable workforce and apposite household size was control of marriage (Doveri 2000). Sons remained in the parental home after marriage to ensure the present and future workforce (through reproduction), but daughters had to leave.

In being landless, sharecropping households had nothing to hand down but their intrinsic and highly organized capacity of work. This induced them to think and behave as a single unit, with each member fulfilling a specific role and with particular duties that cemented ties among all household members. Many authors have stressed the complexity of these roles and internal relationships (Grilli and Zanotelli 2010; Micheli 2011), which prove to be far less idyllic than might appear, as noted by Das Gupta (1999) for stem and joint families in Asia, and by Cavallo (1998:93) for Italian complex households, which were often characterized by “conflictual relationships and power inequalities.” Within this family system, the elderly were vulnerable from a biological point of view but not from a social one. The strictly hierarchical sharecropping family structure meant that the household head was typically the senior male. This figure held unquestionable decisional power over the entire family group, management of labor, and running of the estate. He alone was authorized to sign contracts and communicate with the landlord, approve the timing of marriages, decide whether and when someone had to leave, and decide whether to hire someone. In addition, because no formal retirement existed, the household head retained this position until death.

However, as many scholars point out, all theoretical frameworks addressed at defining a specific family form and family formation system for Italy have been largely unsuccessful. As Kertzer put it, Italy is “a burial ground for many of the most ambitious, and well known, theories of household and marriage systems” (1993:247). This is also true on a local level, where systems varied according to socioeconomic status, ties with the land, and/or occupation. In Tuscany, the marriage and family formation system as well as living arrangements of poor day laborers were completely different than, if not opposite of, sharecroppers. In fact, the day laborer group has certain characteristics that are much more typical of Northern European populations, such as low permanent celibacy rates, neolocalism, and nuclear-based family systems. Although sharecroppers and day laborers were both landless, the latter also lacked direct access to land, living not on the farm but in the village. They were usually hired on a daily basis by smallholders, farm tenants, or sharecroppers when the demand for agricultural workers was high (typically from mid-spring to autumn). When this demand fell, they dedicated themselves to modest handicrafts or moved away in search of work elsewhere.

The precarious work of day laborers did not involve the whole family group, meaning that the household was not a production unit. Household structure did not, therefore, overtly determine finding or maintaining employment. In fact, large and complex households could become an untenable economic burden or particularly problematical in the event of moving from village to village. This explains why day laborers adopted neolocal patterns of living arrangement upon marriage and lived in small and simple households. Reher (1998) described this residential pattern as a weak-tie family system, with the elderly living in isolation from married sons and relatives.

The Community of Casalguidi

Casalguidi is located in the present-day province of Pistoia, in Tuscany, about 30 km north of Florence and 7 km south of Pistoia. In the period under study (1819–1859), the parish belonged to the Grand Duchy of Tuscany and was the most populous of the Pistoia area, with an average of around 2,400 inhabitants per year. There, as elsewhere in rural Tuscany, agriculture was by far the most important productive and economic sector: grain and other cereals were mostly cultivated in the plain, whereas vineyards and olive groves characterized both the plain and low-lying hills (Chiti 1988). About 80 % of households were involved in agricultural work, resulting in generally low levels of well-being. More than 50 % of households fell into the minimum tax bracket, 35 % were exempt from payment for manifest indigence, and about 15 % were subject to medium/high taxes (Manfredini and Breschi 2008). Even among homeowners, 2 in 10 households were exempt from taxes, which is a sign not only of the modest value of their property but, above all, the difficulty they faced in providing for their own upkeep. In rural areas, as in the rest of Tuscany, most of the population lived in poverty. A sudden increase in living costs, rise in debt to the landowner, serious health problems, or (worse still) the premature death of the household head were all capable of sending a household into economic crisis (Breschi et al. 2010). The low life expectancy at birth for both males and females—about 35 years—is further evidence of the meager living standards and economic conditions of the Casalguidi population.

The family formation systems and household typologies present in this area (Table 1) partially recall those described earlier for Tuscany. Both sharecroppers and day laborers were characterized by late marriage, although the former had far higher levels of celibacy. These marriage and celibacy patterns were associated with a patrilocal living arrangement for sharecroppers and a neolocal one for day laborers. However, forms of multiple households were much less widespread (29 %) than in other Italian sharecropping areas, where it was not uncommon to find proportions of about 70 % (Kertzer 1989). Joint family units were even fewer, accounting for just about 8 % of total sharecropping households. These peculiarities were mainly due to the combination of two elements. First, the territory of Casalguidi was characterized by small plots of land, which could be as large as one-half the size of those in other Tuscan sharecropping areas (Biagioli 1975; Contrucci 1839; Pazzagli 1992). The express relationship between the size of the plot and sharecropping household meant that Casalguidi’s farms were unfit for large family groups, which would appear to explain the diffusion of nuclear households even among sharecroppers and the relatively low proportion of multiple households and joint families. Second, demographic constraints, such as high mortality and low life expectancy, could also have prevented the formation of complex joint families (Berkner 1975; Levy 1965).

Sources and Methodology

This study relies on data taken from parish registers, or rather the well-known Catholic registers of baptisms, burials, and marriages integrated with information taken from the Status Animarum (Register of Souls) from 1819 to 1859, a few years before the establishment of the official Italian population register. The Status Animarum was a sort of census recorded by the parish priest during his customary Easter visit to all resident families. For each household member, including servants and non-kin, he took note of the name, surname, age, marital status, and relationship to the head and, in some cases, he collected information on the homeowner.

The continuity over time of the Status Animarum for Casalguidi is good, with only one year missing for the entire period, which allows for an accurate account of the evolution of households and their structures. Moreover, the cross-checking of information between vital registers and Status Animarum has enabled the construction of a new meta-source with individual and household longitudinal data, thereby improving the potentiality of the analysis (Manfredini 1996). As for migration, no direct information is available, so migration has been indirectly assumed as the disappearing from Status Animarum of an individual who was not recorded in the death register.⁷ These data allow us to trace the life course of a household in the light of individual experiences of its members and to study the influence of household composition on demographic events (death, birth, marriage, and migration).

Information on the annual family tax of each household has also been retrieved and integrated with the Status Animarum, which makes it possible to trace the economic status and living conditions of each household and, in turn, of each of its members. Because the number of tax categories changed during the period considered, we grouped them into three homogeneous categories (Manfredini and Breschi 2008): untaxed (those exempt from payment for manifest poverty), low tax (those paying only a small due), and medium/high tax (the wealthiest families).

In the final part of this article, the analysis of old-age mortality is carried out using event history analysis, one of the best and most powerful statistical tools for dealing with longitudinal data (Trussell and Guinnane 1993). A discrete-time approach has been used because the data set is characterized by annual repeated observations. This technique is based on logistic regression, with the dependent variable indicating whether an elderly individual died in any given year. The explanatory variables included specifically address the household context in which the individual lived over time. For categorical variables, the estimated coefficients indicate the risk of death relative to a reference category for each included covariate. For quantitative and continuous covariates, the coefficients express the relative risk for each unitary change of the covariate.

The Elderly and Household Structure

A total of 1,026 people (502 males and 524 females) aged 60 and older were recorded as living in Casalguidi between 1819 and 1859, producing a total of 8,083 person-years.⁸ The average proportion of elderly (both male and female) in the population annually was about 8 %. As for old-age mortality patterns, at age 60, women still had an average of 15 years to live—one more year than men—but this difference widened with age, leveling off at about 1.5 years.

In examining the forms of living arrangements of the elderly, we adopted two analytical perspectives. First, we used cross-sectional data from the Status Animarum to investigate which type of household structure the elderly lived in from year to year. The majority of the elderly lived in complex (about 64 %) and, in particular, multiple households (53 %), whereas a minority lived in nuclear families (31 %).⁹ One-person households were decidedly rare (less than 4 %). Again, compared with other Western European contexts (Gruber and Szoltysek 2012; Ruggles 2010), the form of living arrangement in Central Italy, characterized by a strong-tie family system and consequently high proportion of the elderly living in multigenerational households, stands out for its peculiarity. Second, from an individual perspective, we reconstructed the later-life histories of individuals, tracing the living arrangements each of them experienced during his/her residence in Casalguidi. These results reinforce the findings of the previous analysis. Approximately 75 % of people aged 60 and older had lived at least one year in a complex household, and about 64 % lived in a multiple household. Thus, despite the limited presence of multiple households based on cross-sectional observations (Table 1), living in intergenerational family groups for a time in old age was the norm. It also suggests that extended, stem, and joint households could be different facets of the same family cycle. Structural changes in the living arrangements of the elderly were relatively common; among those observed for at least five years, 54 % had lived in at least two different household structure types, and 13 % lived in at least three. Even for the 75 individuals who lived alone for at least one year, solitude largely represented a temporary condition. Although only 10 of them died and 13 remained solitary, as many as 35 of them emigrated, probably to a son’s household elsewhere; 17 were (re)incorporated into another of Casalguidi’s households.

Clearly, whatever the living arrangement, the role of household head was predominantly filled by an elderly man. At the age of 35, 50 % of men already held this position of power, which increases and remains at more than 80 % after age 50. The inferior social position of women in this context made their chances of headship extremely slim, with the exception of the very last phase of life (age 70 or older), when about 40 % of women of that age were household head. In short, the role of elder males in the family system in Casalguidi was clearly paramount. Marked differences in the headship rate also emerge in relation to household type. Given that, by definition, men in nuclear families became household heads upon marriage, this role was already held by 90 % of these men at age 35, approaching 100 % with increasing age. Furthermore, this position was invariably held until death. Headship in multiple households, however, typically represented greater status and power. Here, the elderly “were provided with continued work roles . . . that placed them in a position of authority over their married children” (Kertzer 1995:368). Men living in these households usually reached headship at a much later stage in life; 46 % of these men were not yet head between the ages of 50 and 54, which falls to 10 % for the age bracket of 70–74 years. These findings make unlikely a role of nuclear reincorporation of elderly males within a son’s household in the formation of complex households (Hareven 1994; Kertzer 1995; Ruggles 2007, 2009). Indeed, only 1 % of elderly men joined a child’s household to form a complex family group compared with more than 12 % of elderly women. The especially high levels of older male headship, especially in complex and multigenerational households, suggest that these households were the product of an intergenerational interdependence of mutual benefit to both fathers and sons in nineteenth century Tuscan rural society. The presence of two or more generations under the same roof, with married sons remaining in the parental home, was a key element in securing a sharecropper family’s future workforce, on which their contractual position, access to land, and socioeconomic status depended.¹⁰

As we note earlier, women were extremely unlikely to become household heads at any age. Inheritance issues and cultural values about the role of women in rural society would have prevented them from assuming full responsibility for household life, proven by the fact that more than 92 % of female household heads were widowed. Women assumed this role almost exclusively in the absence of viable male alternatives. Moreover, this arrangement often represented a temporary solution. Unable to secure good contracts and unsuitable for heavy farm labor, the definitive option for widows in this position was likely to be nuclear (re)incorporation into an adult child’s household.

The Socioeconomic Status of the Elderly

The economic conditions of farmers and peasants living in Casalguidi were not necessarily associated with household head’s occupation (Manfredini and Breschi 2008). In a general context of hardship and poverty, data on family tax—which proves to be a better proxy of household living conditions—reveal that elderly people living in multiple and extended households experienced a better economic situation than those living in nuclear and one-person households. The poorest elderly individuals, living in households exempted from paying taxes for manifest indigence, accounted for 20 % of all the elderly living in complex family groups and 35 % of those living in nuclear households (Table 2). This proportion is definitely higher among one-parent households and solitaries, where it reaches, respectively, 56 % and 90 %. On the other hand, the proportion of wealthy elderly people (with medium/high tax) living in complex households is more than twice that in one-parent and nuclear households. It reflects a general socioeconomic differential between complex and simple family groups, although the gap is here more accentuated. To some extent, this evidence recalls Alter’s (1996) findings of a positive relationship between wealth and household complexity. Therefore, old people living in complex households not only held a position of power, surrounded by a large support network of kin, but also benefited from more favorable economic conditions than the majority of those in nuclear households.

Survival of the Elderly According to Living Arrangement

We have seen that it is possible to discriminate between the elderly living in complex as opposed to nuclear households in terms of family support and economic conditions. It follows that elderly living in coresidence with their kin, and in a relatively good economic situation, were likely to enjoy higher levels of security and general well-being compared with those in poor, nuclear households. This prompts the question of whether this translated into an advantage in terms of survival and lower mortality rates. To test for this hypothesis, we used event history analysis. We ran a series of logistic models to assess the effect of household structure and wealth on the mortality risk of individuals aged 60 and older (servants and clerics excluded).¹¹ Four main categories of household structure were used. Nuclear households were subdivided into two groups: namely, nuclear families with both parents coresiding (reference category) and one-parent households. This choice stems from considerations concerning the supposedly higher mortality of widowed individuals with respect to married ones (Alter et al. 2007; Neven 1998; Nystedt 2002), as well as the typology itself, which could be an intermediate form between a nuclear and stem family. Complex households were split into two subgroups according to their level of complexity: namely, those with only one conjugal unit (extended family groups) and those with two or more conjugal units (joint households and frereches). This information indicates, for each person-year, the household structure of the individual’s residence. One-person households were excluded from the analysis because of collinearity with other variables (such as household head), as were households without any defined structure, given the low number of person-years in this category (116).¹² It should be noted that coresidential kin were not the only potential caregivers of the elderly. Older couples or individuals who lived alone could receive support from married children or relatives living nearby, in much the same way as from coresiding kin. Quoting Laslett (1988), Kertzer et al. observe that “relatives outside the household . . . might well have been of more consequence to individuals in Western European societies with their nuclear family households, than to those who lived in multi-nuclear households” (1992:106). More recently, Micheli (2011) argued that strong family systems, especially those in Southern Italy, were based on a kinship federation of family units rather than the coresidence of kin. For this reason, we introduced a dummy variable indicating the presence of at least one kindred household in Casalguidi: households including siblings and/or children of the individual. Family tax was used as proxy of the household’s overall level of wealth (see the previous section). Individual’s age and position within the household (household head or not) were added to the models as control variables. Separate gender-specific models were devised to take into account the very different social positions of men and women within and outside the household, as well as their different mortality patterns. Last, we introduced a variable relating to the cholera epidemic that had a large impact on the village as well as on the entire region in 1855 (Manfredini 2003). This variable was aimed not only at capturing the effects of the epidemic on the elderly population but also at providing, through analysis of the interaction between cholera and number of children <12 years (Model 2), a preliminary answer to the introductory hypothesis put forward concerning the potentially higher mortality rates in complex households with a greater density of members and larger presence of children.

As Table 3 illustrates, no significant relationship emerges between living arrangement and elderly mortality from Model 1 (M1). Although household structure does not appear to have any direct effect on the survival rates of the elderly, the same cannot be said for socioeconomic status and wealth. It is striking that the risk of dying for older males living in households in the low-tax and medium-/high-tax groups was 70 % lower than for those in untaxed households.

The same pattern can be observed for females, with those in the low-tax and medium/high-tax groups having a 53 % and 44 % lower risk, respectively, than those in the untaxed group. However, an exception emerges for older women living in one-parent households, who presented a counterintuitive and significantly lower risk of dying compared with those in nuclear settings (−63 % in M1). This exception can be interpreted in two ways. First, the majority of these households, which were formed by widows with one or two never-married adult children, were transitory forms destined to change as soon as one or more children wedded and/or moved out. The time spent in this family situation was too short to have any effect on the survival of elderly women.¹³ Second, it is possible that elderly widowed women have some advantage in living with children and without a husband: less involved in caring practices and no longer in a submissive position, they could experience an improved quality of life.

The 1855 cholera epidemic had a dramatic impact on the elderly population of Casalguidi, more than doubling the risk of death for women and doubling it for men. However, no statistically significant effect emerges (see likelihood ratio tests) when the variable for the epidemic is interacted with household structure.

Because of the limited effect of household structure alone, we shifted our analysis from a typological categorization of family forms to a more dynamic examination of their members (Table 4). The first model includes the household sex ratio (Males/Females) and a producer/consumer index (Alter et al. 2000).¹⁴ Both indicators were subdivided into three categories (low, average, and high) to allow for a nonlinear relationship with the risk of death.¹⁵ In a way, it is an extension of the Chayanov’s theory of peasant household microeconomy and resource allocation, which postulates that the well-being of household members, especially the weakest ones, is strictly dependent on the household’s capacity to guarantee sufficient resources through its working capacity, which in turn determines the degree of vulnerability of the whole household (Chayanov 1989). On the one hand, households with high values in the producer/consumer index and high values for the sex ratio would have had, in theory, a greater capacity to provide the elderly with resources, maintain or improve living standards, and secure better contracts. On the other hand, households with low values for the sex ratio, indicating a strong female presence, would have meant a greater number of caregivers for the elderly. For this reason, sex ratio calculations excluded children under 8 years of age. As Table 4 shows, our results suggest a prevalence of the second factor.

The sex ratio emerges as much more important than the producer/consumer index in shaping elderly mortality, especially for women. In fact, whereas no significant changes surface in the overall model in the risk of death in old age when associated with the producer/consumer index, households with low values in the sex ratio show a 32 % lower risk. However, it remains unclear whether it was the low presence of males or the high presence of females that contributed most directly to this effect. We therefore ran an additional model (Table 5) estimating the risk of death for the elderly associated with various kin members, which should make it possible to assess the forces within complex households that, in counterbalancing each other, may explain the absence of effect that emerges in Table 3. In a previous comparative paper, Casalguidi showed a negative association between elderly mortality and presence of children (Tsuya and Nystedt 2004). Here, we shift the focus on the number rather than on the presence of kin members. This model estimates the relative effect of the number of coresiding children under the age of 12, unmarried and never-married males and females (12–59 years), and individuals aged 60 and older. The number of unmarried males could have some significance in the context of nuclear sharecropping households, who occupied the small-scale farms in the area. In this context, the possible benefits for an elderly sharecropper of living with a large number of unmarried sons could well offset the supposed advantage of complex households. Introducing the variable concerning the presence of children younger than age 12 makes it possible to test the hypothesis that they represented a health threat to the elderly by serving as natural vehicles of infectious disease (House and Keeling 2009; Marathe et al. 2011). And last, a growing body of literature identifies adult daughters, along with their spouses, as the principal caregivers for the elderly (Hareven 1994)—a factor that the previous classification of household typology did not allow for. Calculations do not include the elderly individual.

Because the results concerning nonhousehold variables are consistent with those for the previous models, our focus here is solely on household-related factors. The overall model shows a significant negative effect associated with the presence of the frailest (the youngest and the elderly) males. In fact, the risk of death of elderly individuals increases in direct relation to the number of young boys and elderly men residing in the household. Conversely, female presence is generally associated with lower risks of death for the elderly. The presence of older women emerges as a particularly protective factor for the survival of elderly men, reducing the risk of death by as much as 35 % with every unitary increase in number. These results further explain the role of the sex ratio that came to light in the previous model. It appears to have been the limited male presence within households—or, more accurately, of certain male figures—rather than the female component that had the most direct influence on the survival rates of the elderly. A possible interpretation of these results relies on the concepts of domestic competition and social hierarchy. In general, elderly individuals, with the exception of the household head, could have been disadvantaged by the co-presence of other nonproductive or less-productive members because allocation of resources tended to favor productive adult members. Given these premises, it is not surprising to find a significant negative impact of the presence of elderly men on elderly women—or at least a stronger effect than on elderly men. As a result of their lower social position within the household and greater dedication to domestic life, elderly women would have experienced more vulnerability to internal competition with other males than would elderly men. In this situation, therefore, competition for resources coupled with social hierarchy and social segregation. That elderly women were significantly (53 %) more likely to die if married and that, all other factors being equal, their chances of survival improved upon widowhood clearly indicate the tough demands that this rural society placed on women, who were considered socially inferior to men and whose main duties were serving the needs of others (primarily husbands, children, and frail family members). Hence, this caregiver role negatively affected the survival of elderly women (as also documented in contemporary societies; see, e.g., Christakis and Allison 2009; Schulz and Beach 1999), but lowered the risk of mortality for elderly men. Not only were elderly women directly involved in assisting sick members, but they were also concerned in daily practices (e.g., the cleaning of surfaces, careful food handling and food preparation, and safe disposal of wastes) that could improve home hygiene and sanitation and therefore limit the spread of infectious diseases. In short, while older men benefitted from a female presence, older women were disadvantaged by a male presence. These findings are consistent with a number of studies on Western societies that also found that widowed women were less likely to die than married women, especially in contexts where gender segregation in occupation and domestic works were stronger and more widespread (Davidson 2001; Hahn 2002; Hughes and Gove 1981; Pizzetti et al. 2005).

The presence of boys under age 12 in the household had a significant negative effect on the survival of elderly men and women. Young boys would undoubtedly have been not only a drain on household resources, because of their importance in guaranteeing the biological continuity of the family group as well as their future workforce, but also competitors to the elderly for care and assistance from the other household members. As for the hypothesis of children as vehicles of infective diseases within the household, the results of the last model do not support this supposition (Table 6).

Conclusions

Many theories on family systems and forms of living arrangements have addressed the idea that complex households have historically played a protective role on the lives of elderly people. The general idea behind this assumption is that large households could provide the elderly with help and assistance more easily than simple family groups could, especially in those societies where public assistance for the elderly was not widespread. The present study, based on a mid–nineteenth century Tuscan rural community, suggests that this was not true in terms of differential mortality. Consistent with Alter’s (1996) findings, elderly people living in complex households were generally in better socioeconomic conditions than those living in a nuclear family; but after we controlled for household wealth, the hypothesized protective role of complex structures did not emerge. Some household members, such as elderly women, emerged as caregivers for the elderly, but other household members were competitors not only for material resources but also for the time and energy dedicated to them by potential caregivers. In this way, the effectiveness of caregiving could be impeded by the number—not the mere presence or absence—of household members needing care and attention, such as boys and elderly men.

What previous theories did not consider is the possibility that the complexity of relationships within complex households may also involve competition and conflict among the members (Oris and Ochiai 2002). Hammel (2005), in his revisitation of Chayanov’s model, maintained that complex households were better able than nuclear households to smooth the negative effects of cyclical variations in the consumer/producer ratio but that such an advantage may be threatened by the higher level of conflict in large family groups. In our opinion, this is exactly what happened for the elderly of Casalguidi. Competition and conflicts among household subunits, as well as general tensions among members of complex households, are negative elements that may have offset the positive effects associated with caregiving and working capacity.

However, the effects of caregiving and competition on elderly mortality in complex households likely depend on many factors, such as cultural values, inheritance rules, household head’s authority and power, existence of retirement, degree of division of work and roles within the household, and gender inequality. In the Chinese population studied by Campbell and Lee (1996), for instance, no effect of coresident kin on elderly mortality was detected. Although this Chinese society was characterized by a joint family system, just like the sharecropping community of Casalguidi, they may have divided the care of the elderly family members among the whole family group; in contrast, in Casalguidi, specific kin (usually women) were more responsible than others for assisting elderly family members.

Profound differences between the urban working-class society of Verviers and the poor rural, family-run economy of Casalguidi were also at the root of the opposite effects observed for the nuclear reincorporation effect. In the Belgium city, married children approximately 40 years old could welcome their widowed parent because at that age, these had (on average) two children of their own who were already contributing to the family income. Conversely, the peasant families of Casalguidi were likely to suffer an economic burden from the arrival of an elderly and often-ill parent (with or without other relatives) because of contractual constraints that tied sharecropping household members. In Casalguidi, this reincorporation could therefore cause a lower risk of old-age mortality in certain forms of living arrangements (such as one-parent or one-person households) because of their intrinsically temporary nature and, conversely, induce a higher risk in the resulting complex households just because of the arrival of elderly people.

Finally, an explanatory element that should not be overlooked is a possible mortality crossover (Manton et al. 1981; Vaupel and Yashin 1985). In a population characterized by heterogeneous subpopulations (Keyfitz and Littman 1980), the frailer individuals of the more disadvantaged subpopulation might be rapidly eliminated, leaving only the more robust ones. Analogously, previous studies have demonstrated that nuclear households experienced higher infant mortality than did complex ones (Breschi et al. 1999, 2004). This stronger selection in the disadvantaged component of the population would produce a faster elimination of frailer individuals in nuclear households. This selection of the more robust individuals at old ages could have then counterbalanced the advantages of individuals living in more protective but less selective contexts.

In the end, flexibility in the composition of a family group was common, with rapid reactions and adaptations to both external and internal stressors. When necessary, fissions, fusions, and changes in the number of members and structure of households were used by mid–nineteenth century Tuscan peasants, and likely in many other populations, in their attempts to manipulate their socioeconomic status or the balance between resources and consumption. This is precisely the reason why only studies based on a life-course analysis of family groups can investigate in depth the effect that family forms and structure might have had on the lives of their members.

Notes

References