Improved life expectancy in France, although accompanied by a decline in the proportion of years lived without disability (Robine et al., 1994), is responsible for an increase in the number of dependent elderly people that is due to accentuate with the aging of the baby-boom generations (Désesquelles, 1999). This prognosis raises the question of care provision for cases of dependency, which in turn leads to the familiar choice between continued living at home and “placement” in an institution. Among more than twelve million people aged 60 and over enumerated in France in the 1999 Census (Courson and Madinier, 2000), close to 500,000 lived in medical and social institutions [1]. Deterioration in health status plays a decisive role in placement in an institution, but other factors are certainly also involved. Foremost among these are insufficient available help, whether professional or informal; inadequate resources; and unsuitable housing (Metzger et al., 1997; Simon and Fronteau, 1999). Given that assistance to dependent persons is very often provided by a family member (Renaut et al., 1995; Dutheil, 2001), we expect that people without families able to provide this assistance, for whatever reason (lack of family or broken family ties, geographical distance, incompatibility with work, etc.), are more likely, other things being equal, to reside in institutions.

The results of the Disability, Functional Limitations, Dependency survey (Handicaps-Incapacités-Dépendance, or HID survey for short) conducted by the French National Institute of Statistics (INSEE) (Mormiche, 1998) make it possible to verify this hypothesis. In 1998, close to 15,000 people living in medical and social institutions were interviewed about their disabilities and about their family environment. One year later, slightly over 17,000 people living in private households answered the same questionnaire (see Appendix). We thus have information with which to describe and compare the family networks of people aged 60 and over depending on their living arrangements (household or institution). The information gathered by this survey, however, led us to adopt a restrictive definition of family: the data are limited to the respondent’s partner [2], children and grandchildren, siblings, and parents and grandparents. We know nothing about the children-in-law, siblings-in-law, or great-grandchildren that respondents may have.

Finally, it should be noted that the respondents were questioned about their living descendants, ascendants, and siblings. The family network to be described is therefore the possible locus of more or less frequent exchanges (visits, telephone conversations, correspondence, various kinds of help). To study the impact of the network on the respondents’ living arrangements, we will first describe the “potential” networks of people aged 60 and over, and then examine the networks in action, again comparing people living in private households with those living in institutions.

Table 1
Distribution by marital status of people aged 60 and over according to living arrangement (in %)

Table 2 indicates the proportion of people living in an institution in each ten-year age group, by sex and legal marital status. Relative to married people, this population has a clear over-representation of widowed, separated, divorced, and most notably single people. Fewer than 1% of married people live in institutions as against 13% of single people, 9% of widows and widowers, and 5% of separated or divorced people. The increase with age in the proportion of people living in institutions (from 1.1% of men and 0.8% of women aged 60-69 to 20.8% of men and 34.2% of women aged 90 and over) is observed regardless of marital status. Finally, whereas men up to age 80, whether single, widowed, or married, are more likely than women to live in institutions, the situation is reversed at the oldest ages: 18.4% of women aged 80 and over live in institutions compared with 9.7% of men of the same age.

Table 2
Proportion of people living in institutions by marital status, age, and sex (in %)

How can the under-representation of married people in institutions be explained? Assuming that the onset or aggravation of dependency plays a role in the move into institutional accommodation, we might ask whether the prevalence of dependency is higher among people who are “single” than among married people. As in the study of differential mortality by marital status (Vallin and Nizard, 1977; Colin, 1996), two hypotheses can be put forward to explain the differential risk of dependency by marital status. The first is a selection effect. According to this hypothesis, the existence or absence of disability explains, at least partly, the marital status of individuals, with disabled people who have not married because of their disability being the most eloquent illustration of this effect. In the case of widows and widowers, we can imagine a selection effect operating as follows. The risks of death and of experiencing disability are known to increase when moving down the social hierarchy (Desplanques, 1993; Mesrin, 1999; Cambois, 1999). Consequently, the probability that one member of a couple will be widowed and experience disability is higher if the partners are manual workers than if they are in professional/managerial positions. The second hypothesis involves a protective effect of marriage. The literature contains various examples of beneficial effects attributable to the presence of a partner (control exercised by the partner on behaviour, in particular alcohol and tobacco consumption and eating habits; encouragement to seek medical care; moral and emotional support), all of which clearly have implications for the onset of disability.

Table 3 traces the prevalence of dependency (of physical and/or mental origin) by age, sex, and marital status.

Table 3
Proportion of people with dependencies⁽¹⁾ by age, sex, and marital status (in %)

Physically dependent persons were defined as:

Mental dependency was determined from responses to the following questions:

People who were totally incoherent or always disoriented, and people who were partially incoherent and occasionally disoriented, were counted as mentally dependent.

Among people aged 60 and over, 14.1% of married women are dependent as compared to 21.2% of separated or divorced women, 23.3% of single women and 30.5% of widows. Among men, widowers (29.5%) and single men (21%) are also more likely than married men (17%) to be dependent, though separated and divorced men are the least affected by dependency (11.4%). The large differences in age structure across marital statuses could partly explain these results. When age is taken into account, the picture that emerges is much less clear than at first appeared. For both men and women, the advantage of married people relative to the other categories as regards dependency is not observed in every age group.

Even if the probabilities of dependency were identical for married and unmarried people, it is not hard to imagine how the presence of a spouse can facilitate maintenance at home once dependency occurs, provided that the spouse himself or herself is in good health and thus able to perform the role of carer. We would moreover expect the effect of marriage to fade with age: the probability of the carer spouse in turn becoming dependent increases as the couple ages. This is shown in Figure 1: for both men and women, the relative risk of living in an institution decreases with age for single and widowed people relative to married people.

Slightly more than half of the people living in institutions have no siblings alive (Table 4) [4], compared with only 27% of people living in private households. But the population living in institutions is older than the population living in private households: the difference between the average age in these two groups is over ten years (83.2 years for the former as against 71.5 for the latter). To control for this bias, we calculated what the proportion of people with no live siblings would be in the institutionalized population if it had the same age structure as the population aged 60 and over living in private households (standardization). The disparity is then significantly reduced (from 54% to 35%), but the difference in age and sex structure between the two populations does not completely explain the disparity that was initially observed. In each age group, the average number of live siblings among people resident in institutions is less than that of people in private households (Table 5). People aged 60 and over in private households have an average of 1.9 siblings alive, while for institutionalized people in the same age group the figure is only 0.9.

Table 4
Proportion of people aged 60 and over with no live siblings, by living arrangement

Table 5
Average number of live siblings by age and living arrangement

Examining the descendants of people aged 60 and older yields similar conclusions. In private households, the proportion of people with no children alive is only 14%, compared with 40% in institutions (Table 6). This gap may also be the result of compositional differences by age and marital status between the two populations. If the institutionalized population aged 60 and over had the same structure by age and marital status as the population living in private households, the proportion of people in institutions with no living children would be only 23%, but it would still be higher than that observed in private households. In each age group, the average number of living children, as well as the average number of living children and grandchildren, is significantly higher in households than in institutions (Table 7 [5]). The difference appears particularly marked in the youngest age groups, where the over-representation of single people in institutions is greatest. We note also that in private households, the average number of living children decreases with age. This change results partly from the increase in the probability of dying for the children as, like their parents, they grow older, but also from the rise in completed fertility over the generations in question: women in the 1900 birth cohort had an average of 2.11 children, as opposed to 2.60 for women in the 1930 cohort. This evolution is not observed for the institutionalized population. This contrast is a further consequence of the compositional difference between the two populations in respect of marital status. If single people are excluded from the calculation, the average number of live children of people aged 60 and over falls with age in both households and institutions (Table 8).

Table 6
Proportion of people aged 60 and over with no live children, by living arrangement

Table 7
Average number of live children and grandchildren by age and living arrangement

Table 8
Average number of live children of ever-married people by age and living arrangement

The data on the family circle of people aged 60 and over has been summarized by distinguishing five main types of family situation [6] (see box). In private households, three people out of five have a partner and children (Table 9); with an identical age and sex structure, this would be the case for only one in ten institutionalized people. Having children but no partner is the most common situation in institutions (58% of cases). But the most unfavourable situations as regards the extent of the family network are also over-represented in institutions: one third of people aged 60 and over living in institutions have neither a partner nor descendants, as against only 8% in private households.

Table 9
Family situation of people aged 60 and over by living arrangement (distribution in %)

Regardless of the variable studied (number of siblings, number of children, number of grandchildren), the respondent’s identity is found to have an entirely opposite effect on the results in households and in institutions. In institutions, the situation is as expected: standardizing for age, the family network is larger when respondents answered the questionnaire unaided than when they answered with partial or complete assistance. In private households, the opposite pattern is observed. Because recourse to assistance in responding to the survey was much more common in institutions than in private households, the differences observed earlier probably over-estimate the real differences, but the “disadvantage” of people living in institutions compared with people living in private households cannot be called into question. For a given age group, and with the questionnaire administered in the same way (answered unaided, answered with assistance, answered entirely by a third party), people in institutions have family networks that are very significantly smaller than those of people living in private households. The close [7] family networks of people living in private households comprise 8.7 people on average, compared with 4.6 persons for those living in institutions (Table 10). The difference seems to be even larger among respondents who answered through a third party (9.2 versus 3.6 persons), while, on the other hand, it is slightly smaller when the person selected for interviewing answers unaided (8.6 versus 5.1 persons).

Table 10
Average size of the close family network⁽¹⁾ by identity of the respondent and living arrangement

Table 11
Proportion of people living in institutions by family situation and existence of dependency (in %)

We assume that this categorization adequately captures the degree of severity of the physical dependency. We also introduced into the model a variable characterizing the respondent’s social origin. This variable corresponds to the last occupation of the respondent, or of the partner or former partner if the respondent has never worked.

The results from this model [8] show that variables describing individuals’ family setting have a very significant effect on the probability of living in an institution (Table 12). The strongest effect is that of union status. Other things being equal, people in a union have a much smaller risk of living in an institution than single people (OR = 0.09). The lack of a partner is thus a key factor in entry into an institution. In addition, having children, whether daughters or sons, is highly favourable to continuing to live at home. The effect of siblings or grandchildren is difficult to determine because information on this point was not provided by about one fifth of people living in institutions.

Physical dependency was expected to have a very strong effect on the probability of living in an institution, and the results went beyond our expectations. Physical dependency, whether or not combined with mental dependency, is an important discriminating variable, whose influence appears to rise in line with the increasing severity of the dependency. More surprising is the lack of effect of mental dependency alone. It is also surprising to note that the age effect remained significant even after controlling for dependency. Is this to be interpreted as a generation effect? It should be made clear that although our model includes a large number of variables, it naturally does not claim to account for all the heterogeneity in the population being investigated. In particular, the level of economic resources is not introduced explicitly, and the variable for social origin only partly controls for the socio-economic heterogeneity of the population. It is possible that this distinctive effect of age masks the effect of differences in income.

Table 12
Factors affecting the probability of living in an institution (results from logistic regression)

Social origin, in any case, appears to be a powerful discriminating variable. The analysis contrasts non-manual workers (OR = 8.55), intermediate occupations (OR = 4.81), and manual workers (3.41) with craftsmen, shopkeepers and company managers (OR = 1.44) and farmers (OR = 1.00). Those in professional/managerial jobs and intellectual professions (OR = 2.69) are in an intermediate position. Other things being equal, therefore, the least well-off social categories have a higher risk of being placed in an institution. Economic considerations, however, cannot explain the position of professional/managerial workers and farmers. The observed “hierarchy” seems to correspond more closely to a contrast between former salaried workers and former self-employed workers. In the particular case of farmers, one possible explanatory hypothesis concerns the existence of a stronger solidarity based in community and on the greater geographical proximity of family members.

When all these variables have been controlled for, women have a lower probability of living in an institution than men [9]. The effect of sex on the probability of living in an institution depends on union status: the same analysis limited to people in couples shows no significant difference between men and women. The relative advantage of women over men is stronger, on the other hand, in an analysis limited to single people. Women, traditionally accustomed to carrying out housecare tasks, in all probability cope better than men with living alone.

Respondents to the HID survey who reported having family were first asked to answer the following question: “Among the relatives that you just mentioned, are there any who live with you or with whom you have contact?”. 5% of people aged 60 and over who have family answered this question in the negative, and this proportion was significantly higher in institutions (12%) than in private households (5%) (Tables 13 and 14). By combining information about the existence of family on the one hand and about contact with this family on the other hand, we constructed a measure of isolation (the situation of people who have no family and of those who have a family but have no contact with it). Isolation is much more common in institutions (27%) than in households (7%), and is more common among women (9%) than men (5%). Whereas the frequency of isolation increases with age in private households, rising from 4% at age 60-69 to 10% at age 90 and over, the opposite pattern tends to be observed in institutions. We note in particular that at age 60-69, almost two in five people living in institutions are isolated.

Table 13
Proportion of isolated people⁽¹⁾ by age, sex, and living arrangement (in %)

This leaves the 93% of people aged 60 and over who do have contact with their close family. These people were asked about the frequency with which they saw family members. More specifically, the survey collected information on the respondent’s partner and parents, attention being limited for the rest of the family to the two children and two siblings [10] that the respondent sees most often. We will use the term “encounter” to designate a relative that the respondent has the occasion to see. A monthly (respectively weekly) encounter corresponds to a relative that the respondent sees at least once per month (respectively per week).

15% of people aged 60 and over living in private households cohabit with a close relative other than a partner, and 16% live near to a close relative. The frequency of encounters between the respondent and the family members who live nearby or who cohabit with the respondent is unknown. In the following analyses, the assumption is made that encounters with these people are weekly [11]. The proportion of people living in private households who see at least one member of their family once a week is 85%, and the proportion rises to 89% for monthly encounters. Comparison with the respective values observed in institutions (43% and 57%) thus shows a very marked “disadvantage” of people living in institutions relative to people living in private households (Table 14).

Table 14
Distribution of the population aged 60 and over by frequency of encounters with close family and by living arrangement (in %)

In order to isolate the individual characteristics that most strongly influence the number of monthly encounters, we employed linear regression. The results are reported in Table 15. The analysis included all people who had any family. Most of the control variables in this model were also included in the model described earlier (sex, age, legal marital status, physical or mental dependency, and social origin of the respondent). The size and structure of the family network are described by four variables: number of sons, number of daughters, number of siblings, and existence of grandchildren. We also introduced a three-category variable to characterize the respondent’s living arrangement (private household, long-stay care facility attached to a hospital [12], and other types of institution [13]). Finally, we added a variable that specifies the identity of the person responding to the questionnaire (the respondents themselves with or without help, a third party).

Table 15
Factors affecting the number of monthly encounters with family (results of linear regression)

After controlling for all these variables, the situation of people living in institutions was not significantly different from that of people living in private households. The disadvantage observed earlier of people in institutions thus results from the different structure of these two sub-populations relative to several variables whose effects we are now able to detail.

We showed earlier that family networks are smaller in institutions than in private households. The number of monthly encounters is very sensitive to the number of living children and siblings. The number of daughters has a more marked effect than the number of sons. The existence of grandchildren is also associated with more numerous contacts with family. The coefficients with the largest absolute value in the analysis are those concerning marital status. Other things being equal, married people are more likely than widowed people to see their family on a monthly basis, while people who are single, divorced, or separated have the least contact. Doubtless the fact of counting the spouse among the people seen monthly explains much of the “advantage” of married people. It is reasonable to think that if this were not the case, the difference between widowed and married people would be negligible.

No difference is observed between men and women. On the other hand, the analysis shows a significant negative effect of age on the number of monthly encounters. Does this result reflect a reinforcing of family ties in more recent generations (a cohort effect) or a change in the nature of these ties with increasing age (an age effect)? To support this second hypothesis, it can be noted that in all probability sexagenarians are asked more often than older individuals to look after grandchildren or care for a dependent older parent [14].

The effect of a dependency is also found to be fairly discriminating. Dependent people are less likely than non-dependent individuals to see their families on a monthly basis. A first explanatory hypothesis is that dependent people are abandoned by their families. This could explain why the effect is stronger when the dependency is mental rather than physical, since the family feels more powerless to provide assistance in the former case. It is also possible that the absence of contact with the family predated the dependency. The observed result in this case would attest more to the negative impact on physical and mental health of a lack of encounters with close family. At the same time, it would provide an argument to refute the idea that the onset of dependency in a close relative has the effect of reactivating family ties.

Social origin also has a highly significant influence on the number of monthly encounters. The hierarchy revealed here is very close to that observed in the study of factors influencing the probability of living in an institution (see Table 12). In particular, the contrast between previously self-employed and salaried workers reappears, with the former having more contact than the latter with members of their family.

Finally, the respondent effect that was mentioned earlier is again significant here. Other things being equal, when the questionnaire was completed entirely by a third party, the number of monthly encounters is higher. This situation, as noted earlier, occurs most often in institutions, and in nine out of ten cases the person who responded is a staff member at the institution. It is possible that in many cases this person will not know that a family exists unless its presence is visible, which presupposes that contact is maintained.

With whom do these reported monthly or weekly encounters take place? Most frequently (in more than three in five cases), they involve a woman (Table 16). In private households, less than half of the people encountered on a monthly or weekly basis are aged 60 and over. This proportion is higher in institutions (three out of five contacts), which is quite logical, since the institutionalized population itself is older. The very low representation of the under-25s is explained by the fact that grandchildren are not taken into account. In the majority of cases (one in two cases in private households, three in four in institutions), the people encountered are the respondent’s children. Half of the people aged 60 and over have at least one child that they see on a weekly basis. For siblings, this proportion is much lower but still non-negligible: 16% of people aged 60 and over see a sibling at least once a week. Finally, the idea that geographical proximity facilitates contact is corroborated: in private households, 87% of reported weekly encounters (68% in institutions) take place with people who live in or around the town or city where the respondent lives. This proportion is only 30% for monthly encounters (32% in institutions).

Table 16
Characteristics of family members encountered on a weekly or monthly basis according to living arrangement (distribution in %)

To answer these questions, we employed multiple correspondence analysis (MCA) followed by a hierarchical classification. This analysis concerned only people living in institutions for the elderly [15] and used the following active variables: age at the time of the survey, sex, union status, size of family network (four categories: no family, one to three persons, four to six persons, seven or more persons), dependency, social origin, length of stay in the institution, frequency of contact with family. The questionnaire administered in institutions contained a question on the frequency with which people exchanged news with a family member by letter or by telephone. People who see an institutionalized relative at least once a month, or who see the relative several times a year with at least one exchange of news per month, are considered to have an “active contact” with this relative. This enables us to distinguish, among the people who have no monthly contact with relatives, those who have at least one active contact from those who have none. We took this as representing the minimal frequency of relations between a person in an institution and his or her family circle necessary to consider that person as not isolated.

Finally, the analysis included as explanatory factors two variables describing the conditions of entry into the institution: age at the time of entering the institution and the motive for entering. One in four people living in an institution for the elderly had not entered it for health reasons, though what the reason was is unfortunately not known. These people are in fact much less likely to be dependent than people who entered an institution because of poor health (59% versus 83%). Comparison shows that they have a slight disadvantage as regards the size of their family network (4.4 people on average compared to 4.8 people) and that they have significantly fewer weekly visitors (0.5 on average versus 0.6). Social origin appears not to be discriminating.

The first ten coordinates produced by the multiple correspondence analysis were used to classify individuals into groups that were as homogeneous as possible. The most detailed of the three best partitions obtained by hierarchical classification distinguishes seven classes, each representing between 11% and 18% of the total population resident in institutions for the elderly (Table 17). In the first four classes, entry into the institution is attributed to a health problem more frequently than the average for the total population living in institutions for the elderly. This is the case in particular for class 3, where 86% of people (74% in the total population studied) moved into an institution for health reasons. Women (82% versus 74%) and octogenarians (68% versus 45%) are over-represented in this class. Severely dependent people are likewise over-represented: 48% of people in class 3 (versus 17% in the total population studied) experience both confinement to bed or armchair and mental dependency, while only 2% suffer from no dependency (versus 22%). The family network size of people in this class (on average 2.8 people) is also much reduced. We can make the hypothesis that the combination of an inadequate family network with a severe dependency, such as creates a substantial need for assistance, is what motivates the entry into an institution for a large proportion of the people in this group.

Table 17
Principal characteristics of the categories obtained from hierarchical classification carried out on the population living in institutions for elderly people

Classes 1, 2, and 4, on the other hand, in which the proportion of people who entered institutions for health reasons is substantially larger than average (from 79% for class 1 to 75% for class 4), are characterized by family networks comparable in size to those found in private households. In addition, the proportion of people who have at least one weekly contact is high (97% in class 1 and 55% in classes 2 and 4, as opposed to 40% in the study population as a whole). Clearly, then, these are people who receive considerable attention from their families. These three classes, which together represent nearly half of the institutionalized population, also possess several distinct characteristics.

Class 1 is characterized by having the highest proportion of people with at least one weekly contact. However, only one third of the people in this situation belong to this class. Men (38% as opposed to 26% on average) and people in a couple (35% versus 6%) are over-represented in this group. Class 2, on the other hand, has a higher proportion of women (87%) than the study population as a whole. It is composed essentially of single people (99% of the class) aged 80-89 (86% versus 45% in the whole of the study population). In addition, this group is distinguished by a high proportion of people who have difficulty going out, washing or dressing themselves (72% versus 52%). Manual workers are very significantly over-represented in this class (53% versus 30%). Finally, 99% of class 4 is made up of single people aged 90 and over. These people entered institutions late (at average age 88.7 years versus 80.6). Women are over-represented in this class to an appreciable degree (90% versus 74%), as are dependent persons (94% versus 78%). With an average of six members, these people have a larger family network than the general population of people living in institutions for the elderly. In fact, it is practically identical in size to that of people aged 90 and over living in private households (6.6 people). Craftsmen and shopkeepers (17% versus 10%) as well as farmers (17% versus 11%) are over-represented in this class.

For the next three classes, socio-economic factors very likely played an important role in the entry into an institution. The family network in these classes appears small, and the proportion of people who entered the institution for health reasons is below 75%. The lowest value appears in class 5, where only 57% of people entered the institution because of a health problem. It is also in this class that the frequency of dependency is lowest (38% versus 78% on average). The average age of people in this class is significantly lower than the overall average (81 versus 84 years). The group is composed exclusively of single people whose family networks are not very large (on average four people). Nevertheless, 78% of them (versus 60%) have at least one active contact. The intermediate occupations (28% versus 9%) and non-manual workers (33% versus 18%) are significantly over-represented in this class.

Classes 6 and 7 are characterized by small family sizes (respectively 0.2 and 2.2 people), very much lower than in households, and by a high proportion of people without active contacts. Class 6 contains the largest proportion of people with no family (90% of the class versus 16% of the study population). All the people in this class have no active contacts. Women (82% versus 74%) as well as non-manual workers (31% versus 18%) and intermediate occupations (13% versus 9%) are considerably over-represented. Class 7 is the youngest class (average age 73 years versus 84 years). In 80% of cases (versus 36%), the respondent entered the institution before age 70. The group contains a majority of men (only 38% women). Non-dependent people (39% versus 22%) and manual workers (57% versus 30%) are over-represented in the group.

This “disadvantage” of people living in institutions is one cause of their higher frequency of relational isolation, but other factors are also at work. In particular, age and the existence of a dependency, which as we know is the situation of a large majority of people in institutions, are associated with a lower intensity of relations with close family members.

Overall, more than two out of five people living in institutions, compared with one in ten in private households, either have no close family or have no monthly contact with their family. But in terms of family circle, nearly half of the population living in institutions has characteristics very similar to those observed in private households. The institutionalized population cannot be considered as forming a homogeneous whole, and it is clear that the diversity of individual situations is matched by a wide variety of reasons for entering an institution.