This Article Abstract Full Text (PDF) Services Download to citation manager

Differences Over Time in the Relationship Between Partnership Disruptions and Support in Early Old Age in Britain

¹ Institute of Gerontology, King's College London, England.
² Department of Economic and Social Sciences, University of Molise, Campobasso, Italy.

Address correspondence to Karen Glaser, Institute of Gerontology, Room 6.24, Strand Building, King's College London, The Strand, London, WC2R 2LS, UK. E-mail: karen.glaser{at}kcl.ac.uk

	Abstract

TOP Abstract Methods Results Discussion References

 
Objectives. We examined variations over time in the relationship between past partnership disruption (due to divorce, separation, and death) and present support (coresidence with, help to and from, and contact with children) in early old age in Britain.

Methods. Employing data from the 2001/2 British Household Panel Survey and the 1988/9 Survey of Retirement and Retirement Plans, we investigated whether differences in support over time were due to changes in (a) the composition of the population (e.g., the percentage divorced) or (b) the strength of the covariates (e.g., the effect of divorce).

Results. Our findings showed an increase in the experience of partnership disruption in early old age. Nonetheless, the percentage of people who reported receiving and providing support increased. A comparison of the two points in time showed that there was little change in the effects of partnership disruptions on support in early old age.

Discussion. The fact that the level of support has remained stable or increased over time among this age group despite the increasing prevalence of divorce suggests that the negative effect of partnership disruption on support in early old age may be weakening over time.

Key Words: Partnership dissolution • Older people • Divorce • Support

BRITAIN, like other developed countries, has witnessed significant changes in family structure and relationships in recent decades. Among the well-documented trends are declines in marriage and childbearing and rises in divorce and cohabitation (Haskey, 1993; Stevenson & Wolfers, 2007). Such changes suggest that today's older people have more diverse family networks than previous generations (Wachter, 1997). Nevertheless, it is surprising how little researchers know about the impact of such trends on intergenerational relationships in later life (and on support in particular). On the one hand, there is a popular perception that increasing rejection of traditional family roles means that family members have become less willing to care for one another. This idea is to some extent supported by empirical research. For example, most researchers have considered that the experience of divorce weakens family ties and thereby reduces mutual support, perhaps due to greater geographic mobility, social disapproval from family or friends, or loss of reciprocal obligations (Cherlin, 1992; Ganong & Coleman, 2006). On the other hand, the notion of what "the family" means has changed, with a more flexible and diverse concept gaining ground of "family members," among whom reciprocity and affection (in addition to obligation) are key influences on intergenerational relationships and exchanges (Askham, Ferring, & Lamura, 2007). As notions of "family" life change, with greater acceptance of divorce and increasing tolerance of different family forms, the negative impact of such changes for support in later life may be disappearing (Thornton & Young-DeMarco, 2001).

Given that social networks at older ages largely comprise family members, understanding how trends in family life may affect future support and care is a critical issue (Askham et al., 2007). To our knowledge, no studies have examined changes in the relationship between partnership disruption among older parents and support over time; most have relied on microsimulation methods to estimate the future potential impact of divorce (Wachter, 1997).

Given concerns about future support at older ages, this absence is surprising. In this study, we examined changes over time in the association between partnership disruption and support (coresidence and contact with children, and help given to and received from children) among 61- to 69-year-old mothers and fathers using data from the 2001/2 British Household Panel Survey (BHPS) and the 1988/9 Survey of Retirement and Retirement Plans (RS). The data permitted us to investigate (a) changes in support given and received and (b) the extent to which any differences result from changes in the composition of the population (i.e., the mix of individual attributes associated with the propensity to give or receive support) or changes in the strength of relationship between explanatory variables and support outcomes.

Conceptualizing and Operationalizing Support
One of the difficulties in studying support is its conceptualization and operationalization (see Barrera, 1986; Hermalin, 2002; House, Landis, & Umberson, 1988). Support is usually defined in terms of (a) structural characteristics of the social support network; (b) social embeddedness (e.g., the frequency of contact with others); (c) emotional assistance (reflecting current and future availability of emotional support); and (d) instrumental assistance (e.g., transfers of space, time, and money) from family, friends, neighbors, and other community members (Barrera, 1986; Hermalin, 2002; House et al., 1988; Soldo & Hill, 1993). In the BHPS and RS, comparable support outcomes are (a) coresidence, (b) frequency of contact with children, and (c) money transfers and assistance with household tasks to and from children.

Empirical Evidence on the Relationship Between Family Disruption and Support
Most work on the long-term impact of partnership dissolution has focused on outcomes for children. However, there is growing evidence in the United States and The Netherlands that partnership disruptions over the life course (particularly divorce) have deleterious consequences for support at older ages (Aquilino, 1994; Bulcroft & Bulcroft, 1991; Cooney & Uhlenberg, 1990; Furstenberg, Hoffman, & Shrestha, 1995; Kalmijn, 2007; Pezzin & Schone, 1999; Shapiro, 2003). In comparison to those in intact marriages, those who have divorced and remarried tend to experience decreased contact and relationship quality with adult children as well as perceived support from children (or from any source; Aquilino, 1994; Bulcroft & Bulcroft, 1991; Cooney & Uhlenberg, 1990; Furstenberg et al., 1995; Kalmijn, 2007; Pezzin & Schone, 1999; Shapiro, 2003). Fewer studies have focused on transfers between parents and their adult children, and here the evidence is less clear: Some studies have reported no relationship between parental divorce and any help given or received (Aquilino, 1994; Pezzin & Schone, 1999), whereas others have reported a negative relationship with time and money transfers (Furstenberg et al., 1995; Kalmijn, 2007). Contrary to previous studies, we have found that older separated parents in the United Kingdom are more likely to receive help from children compared with those in a current partnership (Glaser, Stuchbury, Tomassini, & Askham, 2008).

Some studies have also investigated the association between widowhood and support. Unlike parental divorce, most studies have shown no significant relationship between widowhood and contact (Bulcroft & Bulcroft, 1991); however, the few studies able to examine widowers separately have shown lower contact levels compared with fathers in intact marriages (Kalmijn, 2007). With regard to measures of instrumental assistance, most research has shown that widowed parents receive more support from children compared with still-married parents (Ha, Carr, Utz, & Nesse, 2006). However, a recent study showed a negative relationship between widowerhood and instrumental assistance (Kalmijn, 2007).

Studies have established some variations in the long-term outcomes of partnership dissolution and identified influential factors. Parent's gender is clearly important, for partnership dissolution has a greater negative impact on late-life support for men in comparison to women, which is generally explained in terms of mothers being closer than fathers to their children (Furstenberg et al., 1995; Kalmijn, 2007). The existence of stepchildren is a factor, for older parents are less likely to receive assistance from stepchildren than biological children (Pezzin & Schone, 1999). However, despite changing trends in, and attitudes toward, family life, to our knowledge no studies have examined changes in the relationship between partnership disruption and support over time.

	METHODS

TOP Abstract Methods Results Discussion References

 
2001/2 BHPS
The BHPS is an annual longitudinal survey of a nationally representative sample of private households in the United Kingdom (M. F. E. Taylor, Brice, Buck, & Prentice-Lane, 2007; University of Essex, Institute for Social and Economic Research, 2006). The original sample at Wave 1 (1991) was more than 5,000 households containing 9,912 interviewed adults, with a response rate of at least one interview in 69% of eligible households and full interviews with 92% of eligible respondents in responding households (M. F. E. Taylor et al., 2007). Of the original sample members with a full interview for the first wave, around 62% were still interviewed at Waves 11 and 12 (of the rest, around 10% had died since Wave 1).

The original BHPS sample of households was obtained through a stratified clustered design with 250 postcode sectors in England, Scotland, and Wales (M. F. E. Taylor et al., 2007). This is similar to the sample design used for the General Household Survey, the longest running national cross-sectional survey in Britain. The sampling frame used was the Small Users' Postcode Address File, the frame used for most large-scale government surveys in Britain. All 1991 household members were interviewed and are followed individually until they die, move permanently into an institution, or emigrate; at each wave, data are also collected from other current household members of each original sample member. The sample is replenished through births to original sample members.

We based our analyses on survey Waves 11 (2001/2) and 12 (2002/3; the only waves to include measures of support) linked to the partnership and parenthood histories available in earlier waves. Although the sample size was around 10,000 adults for earlier waves, Waves 11 and 12 had sample sizes of between 17,000 to 19,000 due to the inclusion of the European Community Household Panel sample in 1997 and Scottish and Welsh extension samples in 1999 (M. F. E. Taylor et al., 2007).

There are numerous weighting adjustments available within the BHPS. This study used the cross-sectional weights for individual respondents, allowing for the incorporation of the European Community Household Panel subsample as well as the Scottish and Welsh extension samples. These weights also adjust for unequal selection of addresses and for nonresponse at the household level and the individual level within responding households. The weights were rescaled to the raw sample size (for more details on the weighting procedures in the BHPS, see M. F. E. Taylor et al., 2007). Following advice from the BHPS team at the University of Essex, the cross-sectional weights at Wave 11 were assigned to those who were asked the support questions in Wave 11, and the Wave 12 weights assigned to those who were asked the support questions in Wave 12.

1988/9 RS (in Britain)
The RS conducted 3,543 interviews with a nationally representative sample of people aged 55 to 69. Interviews were achieved at 88% of households (2,565) found to contain at least one eligible person (Bone, Gregory, Gill, & Lader, 1992). All household members within the age range were recruited to the study, so the survey contains a high proportion of couples.

Like the BHPS and the General Household Survey, the RS is based on a stratified clustered design with 100 postcode sectors in England, Scotland, and Wales (Bone et al., 1992). As there is no comprehensive list of people by age living in the community in Britain, obtaining a sample of households containing 55- to 69-year-olds required an initial sift. Thus, the 150 addresses sampled from each postcode sector (using the Small Users' Postcode Address File as the sampling frame) were sent an initial screening questionnaire asking for the age and gender of all adult household members (Bone et al., 1992). As nonresponse to the postal questionnaire was quite high (23%), 1 in 4 households who did not respond to the postal sift were subsequently contacted by interviewers (as were all those living in inner London, as it was assumed that nonresponse in this area would be high; Bone et al., 1992).

We used derived weights created by the Institute for Fiscal Studies and the Department of Social Security (see Disney, Grundy, & Johnson, 1997). These 1988/9 cross-sectional grossing factors were kept as close as possible to the original weights created by the then Office for Population Censuses and Surveys (now the Office for National Statistics). As in the original weights, adjustments were made for sift type (postal vs interview), gender and age group in order to take variations in nonresponse into account (Disney et al., 1997). The main difference between the Institute for Fiscal Studies and Office for Population Censuses and Surveys grossing factors was that the weights for nonrespondents to the initial postal sift who were subsequently interviewed were not adjusted (Disney et al., 1997). Thus, the Institute for Fiscal Studies assumed that individuals who responded to the postal sift were no different from those who responded to the interview. We rescaled the weights back to the raw sample size.

Our analyses used the weights described above to correct for differential response rates within the group of selected individuals and households in both surveys as well as to correct for the nonequal probability of selection of respondents due to the additional subsamples in the BHPS (however, all tables show unweighted Ns). As the RS does not contain variables identifying primary sampling units, or the strata used within each primary sampling unit, we were not able to adjust for design effects due to clustering and stratification (rerunning the BHPS models with these adjustments did not significantly affect the standard errors for the partnership variables).

Sample Selection
We based our analyses on people aged 61 to 69, as Wave 12 (2002) in the BHPS only collected information on support among respondents aged 61 and older, and the upper age limit for RS respondents in 1988/9 was 69 years. As the RS is restricted to Britain, the BHPS analytical sample was similarly restricted (excluding Northern Ireland). Sample sizes were 644 mothers and 546 fathers in the BHPS and 937 mothers and 801 fathers in the RS, all aged 61 to 69 years. In addition to comparable support measures, both surveys collected detailed partnership and parenthood histories, permitting the creation of indicators of partnership disruption. Both data sets have been previously used in cross-sectional analyses of changes over time (Berthoud & Gershuny, 2000; Disney et al., 1997). Key characteristics in each have been compared to the other major national data sets in Britain, such as the previously mentioned General Household Survey, and have been found to be comparable (Disney et al., 1997; A. Taylor, 1994).

Dependent Variables
In 2001 the BHPS collected for the first time, in the new aging and retirement module, information on social support networks. This was for the non-extension survey samples only; the module was repeated in 2002 for the Welsh and Scottish extension samples. The RS collected the same data in a similar form in 1988/9. Both surveys collected details of all household members, including their relationships one to another.

Our study focused on three comparable dependent variables in the BHPS and RS: (a) coresidence with children, (b) frequency of seeing the child living outside the household with whom the respondent reported the most contact, and (c) money transfers and assistance with household tasks to and from children living outside the household. Coresidence with children was determined from the household roster in both the BHPS and RS. With respect to frequency of contact and help given and received in the BHPS, any respondent with at least one non-coresident child (whether or not that respondent also had a coresident child) was asked about frequency of contact with the most frequently seen child (coded as daily, at least once a week, at least once per month, several times per year, less often, and never). We created a binary variable distinguishing those seeing a child at least once a week from those who saw him or her less frequently or never; in creating this variable, we assumed that anyone with a coresident child had more than weekly contact, irrespective of any answer that person gave regarding non-coresident children. For those reporting at least one non-coresident child (whether or not they also had a coresident child), questions followed about the type of help given and received: "Nowadays, do you regularly or frequently do/receive any of the things listed on this card for/from your children who are not living here?" Types of help listed included activities such as lifts in car, shopping, and so on. We created two binary variables: One distinguished those who gave any type of help from those who did not, and the other distinguished those who received any type of help from those who did not.

The questions on frequency of contact in the RS were very similar; we created a binary variable similar to that for the BHPS. Questions followed on help given and help received: "Nowadays (do you/ does either of you) regularly or frequently (do/receive help with) any of the things on this card for/from your child(ren) (include step and adopted children)?" The types of help were almost identical to those in the equivalent BHPS question. However, there were several differences between the two surveys. The first was that the help questions in the RS were asked at the couple level so that individual responses within a couple were duplicated (in the BHPS they were asked of each individual). To test the level of bias that this may have introduced, we reanalyzed the BHPS data to examine the consistency of responses between couples (who were both parents and both sample members). We found that the response to the questions on help received from children was fairly consistent when a dichotomous measure was considered (help vs no help). Another difference between the BHPS and the RS was that personal care was listed in the types of help in the BHPS. In the RS, personal care questions were only asked at the end of a long health module: "Does anyone regularly help you with everyday activities like bathing, dressing or getting about?" and if so, "Who usually helps you?" One possible pre-coded source of help was "Son/daughter not in household." We combined responses from these questions with those from the other help questions.

A third difference between the BHPS and the RS was that in the latter questions on help given and received were asked with reference to any child, and it was not possible to distinguish answers that referred to coresident children from those that referred to children living outside the household. Because in the BHPS these questions referred only to children outside the household, in both surveys we limited our analysis of help given and received to parents with no coresident children. This restriction was likely to bias the sample against more familistic families and those with younger children.

Independent Variables
In order to estimate comparable models over time, we developed a set of similar independent variables in each sample. Partnership disruptions were assessed by (a) measures of partnership dissolution (due to death, separation, divorce, and repartnering) and (b) indicators of family composition (i.e., existence of biological, step- and adopted children).

In the BHPS, full partnership and parenthood histories were collected in Waves 2, 11, or 12; shorter question modules were used for some new entrants in Wave 8 onward. We updated partnership histories wave by wave using indicators of partner in the household and questions on changes in marital status. We decided to study partnerships without distinguishing between the legal and the consensual. This was because the unusual richness of data offered by the BHPS made it clear that respondents themselves did not necessarily recognize this distinction (see Stuchbury, Glaser, Tomassini, & Askham, 2005). In the same way, the category "divorced" included all those who reported a divorce or separation regardless of whether they had experienced a legal marriage, divorce, or separation. The 1988/9 RS also collected detailed partnership histories from respondents. From the partnership histories of both surveys we created indicators of partnership disruption distinguishing those who were (a) remarried, (b) widowed, and (c) divorced or separated from those still in their first marriage (the reference category).

Indicators of family composition (including adopted and stepchildren) in the BHPS and RS were derived from the parenthood histories. One limitation in the RS is that currently married men were not asked how many natural or stepchildren they had had, even if they (or their current partner) reported a previous marriage; consequently, we were not able to include this variable in our analyses of fathers. From the parenthood histories we created a binary measure reflecting whether the respondent had ever lived with a stepchild or stepchildren (including children of a cohabiting partner and children of any age, not just minors). Finally, all BHPS and RS sample members, including married men in the RS, were asked how many living children (without further specification) they had in total; this was the measure used in our analysis to distinguish parents from nonparents and for the number of living children variable.

Other covariates included age, education, social class, tenure status, and health. Previous studies have identified all of these socioeconomic characteristics as key determinants of late-life support (Pezzin & Schone, 1999).

Analysis Plan
Our aim was to study changes in the correlates (and in partnership disruption in particular) of provision and receipt of support among older people. The descriptive and bivariate analyses examined differences in the key characteristics of mothers and fathers between the two time points using t tests and chi-square tests as appropriate for the outcome. For the multivariate analyses we used logistic regression, as comparable support outcomes were binary. Following Tomassini and Wolf (2000), we explored whether differences in support over time were due to changes in the (a) mean values of the independent variables (i.e., the percentage experiencing family disruption) or (b) regression coefficients (i.e., the effects of family disruption). We conducted separate analyses for fathers and mothers given well-known gender differences in late-life support (Kalmijn, 2007). Our approach to missing data was to omit those few cases in which data were missing for any of the dependent variables. In the models, no more than 4% to 5% of the observations were missing for any independent variable; we omitted those cases and conducted our analyses on the valid cases remaining.

The models included interactions of all covariates with year of survey: Significant interactions indicate significant differences in coefficients between 1988/9 and 2001/2. Allison (1999) proposed that such an interpretation in logistic regression may lead to invalid conclusions if residual variation differs across groups. Thus, we tested for unequal residual variation using STATA's testnl option in order to address the potential pitfalls in cross-group comparisons of logit coefficients as described by Allison.

	RESULTS

TOP Abstract Methods Results Discussion References

 
Table 1 presents for mothers and fathers the weighted mean values of the explanatory variables used in our analyses (as well as the t-test and chi-square statistics where appropriate for the null hypotheses of no change in these values between 1988/9 and 2001/2). The t-test and chi-square statistics shown in Table 1 indicate that in nearly all cases there were significant differences between the two time points. For example, the percentage of remarried mothers among those aged 61 to 69 significantly increased from 7% to 20% between 1988/9 and 2001/2 (see Table 1). Paradoxically, despite increasing levels of partnership disruption over time, levels of support as measured by coresidence, frequency of contact, or help given or received largely remained unchanged or increased.

The most common type of help given to non-coresident children by fathers (at both time periods) was looking after grandchildren, giving lifts in car, and helping with money. For mothers it was looking after grandchildren, giving lifts in car, and helping with money in 1988/9 and helping with grandchildren, providing or cooking meals, helping with money, and washing in 2001/2 (results not shown). The most common types of help received among mothers and fathers were giving lifts in car; shopping; decorating, gardening, and repairs (results not shown). Clearly some tasks may have lost their importance over time due to changes in technology and availability; for example, the significant reduction in children helping their parents with washing and ironing (which declined from 6% among fathers receiving this type of assistance in 1988/9 to less than 2% in 2001/2) may imply an increased presence in the older parents' homes of washing machines.

Tables 2 and 3 show the results of our multivariate analyses. We report odds ratios for the estimated parameters, 95% confidence intervals, and levels of significance for 1988/9 and 2001/2. We also report whether the parameters differed significantly between 1988/9 and 2001/2 (i.e., tests of the stability of the relationships between individual explanatory variables and support over time). In interpreting the odds ratios, recall that each represents the effects of a given explanatory variable on the odds of coresiding with, contacting, or providing or receiving help from children. When the odds ratio is larger than 1, there is a positive relationship between the explanatory variable and the outcome; when the odds ratio is smaller than 1, a negative association.

Divorced mothers and fathers, in comparison to those still in their first marriages, were less likely to give help to children; however, the relationship was significant only for fathers. Divorced fathers in 1988/9 reported 90% lower odds of providing help to children; however, in 2001/2 the odds ratio was no longer significant (and there were significant differences between the cohorts with respect to these parameters). In terms of receiving help from a non-coresident child, divorced mothers reported 3.6 times the odds of receiving such help in 2001/2 in comparison to those still in their first marriage. Moreover, there were significant differences between the two parameters: Divorced mothers were significantly more likely to receive assistance from non-coresident children in 2001/2 compared to in 1988/9. Although divorced fathers in this age group reported lower odds of receiving help from children, the relationship was not significant.

The other covariates considered (except for age and health, which were negatively related) showed little relationship with providing help to non-coresident children. In contrast, no educational qualifications, manual work, social sector tenancy, and poor health were generally positively associated with help given by non-coresident children for both mothers and fathers.

Table 3 presents the results for the multivariate analysis of coresidence and contact among parents aged 61 to 69. Remarried mothers in 1988/9 reported 54% lower odds of frequent contact with children in comparison to their still-married counterparts. Remarried fathers reported 57% lower odds of coresiding with children in 1988/9, 66% lower odds of contact with children in 1988/9, and 48% lower odds of contact with children in 2001/2. Widowed mothers reported higher odds of both coresidence and contact; however, only in 1988/9 did the parameter for coresidence reach statistical significance. Widowhood among fathers was generally associated with lower odds of coresidence and contact; however, the link was not significant. Divorced mothers generally showed lower odds of both coresidence and contact, but the association was not significant. By contrast, divorced fathers reported significantly lower odds of contact with children in both 1988/9 and 2001/2.

The other significant explanatory variables in the model (except for age) showed a largely positive relationship with coresidence. For example, those with no educational qualifications and manual workers reported increased odds of coresidence with children (the parameters for fathers who were manual workers differed significantly between the two cohorts). In terms of frequency of contact, once again lacking educational qualifications and being a manual worker increased the odds of contact (the parameters for this latter group differed significantly for fathers between the two cohorts).

For the models presented here we used STATA's testnl option to test for unequal residual variation (a potential problem in cross-group comparisons in logistic regression models) and failed to reject the null hypothesis in all of the models. This suggests that any apparent change in effects over time results solely from unequal residual variation. This supports our results above showing few significant changes in the effects of explanatory variables on outcomes over time.

Another way of presenting the results from the logistic regression models is to calculate the probabilities of each category of the dependent variable for selected characteristics of the parents. We created three profiles for each category of the dependent variable: the 1988/9 profile (applying the 1988/9 coefficients to the 1988/9 mean values of the covariates), the changed structure profile (applying the 2001/2 coefficients to the 1988/9 mean values of the covariates in order to capture how the probability of giving/receiving support would have changed in 2001/2 if the structure had remained the same as in 1988/9), and the 2001/2 profile (applying the 2001/2 coefficients to the 2001/2 mean values of the covariates). Figure 1 illustrates the effects of changes in the population composition between 1988/9 and 2001/2: The first and third columns represent the situation in 1988/9 and 2001/2, respectively, whereas the middle column reflects changes in the probability of giving/receiving support due to differences between 1988/9 and 2001/2 in changes in the effects of the characteristics of the older parents in our samples. Figure 1 shows that for both mothers and fathers, increases in the percentage giving or receiving support are largely a result of changes in the effects of the variables over time rather than in the changes in the means. For example, for both fathers and mothers giving or receiving help from children, the larger differences between the first and second columns are due to changes in the effects (as the means of the covariates remain constant at their 1988/9 levels), whereas the smaller differences between the second and third columns reflect differences in the mean characteristics between the two time periods (as the parameters remain constant at their 2001/2 levels). The observed increase in the probabilities of giving and receiving support between the 1988/9 profile and the 2001/2 profile is largely the result of changes in the propensity to give/receive support.

View larger version (31K): [in this window] [in a new window]   Figure 1. Predicted probabilities of support for three profiles of mothers and fathers aged 61 to 69

	DISCUSSION

TOP Abstract Methods Results Discussion References

Although research shows that the stigma associated with divorce has declined, relatively little work has focused on the implications of this (Gerstel, 1987; Thornton & Young-DeMarco, 2001). A recent study that examined the impact of divorce on children over time using two British birth cohort studies (the National Child Development Study and the Birth Cohort Study, capturing those born in 1958 and 1970, respectively) showed largely unchanged negative effects of parental divorce on outcomes for children (Sigle-Rushton, Hobcraft, & Kiernan, 2005). Our analyses allow us to consider two types of changes over time: changes in the structure of support decisions, reflected in changed parameter values in the logistic regression models; and changes in the population composition, reflected in the mean values of the measured attributes. Like Sigle-Rushton and colleagues, we find few statistically significant differences between our two cohorts in the estimated support parameters for divorce, remarriage, and widowhood; however, where there are significant differences these reflect either divorced older mothers being more likely to receive help from non-coresident children or the disappearance of the significant negative effect of divorce on providing help to non-coresident children by divorced older fathers. When we consider changes in the population composition, we find statistically significant increases in the percentages of older mothers and fathers experiencing remarriage and divorce. Holding the population characteristics constant at their 1988/9 levels but applying the 2001/2 coefficients results in few differences in the predicted probabilities of support between 1988/9 and 2001/2; the larger differences lie in comparing the predicted probabilities that hold the 1988/9 means constant but change the regression coefficients (see Figure 1). The fact that the prevalence of divorce and remarriage has significantly increased, that there have been no significant changes in the effects of partnership disruption, while at the same time the level of support has remained stable or increased suggests that other factors are at work. This raises the possibility that the negative effect of partnership disruption on support is weakening over time; perhaps such events are becoming less crucial in shaping relations between older parents and their children. As demographic and social changes have led to an increase in the number of years adult children and parents co-survive, adjustments to the parents' partnership disruption are likely (e.g., in the attribution of blame for the parental marriage breakdown; Soldo & Hill, 1993). Research has shown that earlier parental behavior (i.e., responsive vs restrictive) had no relationship for fathers with willingness to ask for help with personal or financial problems in later life (Schooler, Revell, & Caplan, 2007).

Moreover, our own previous research suggests that in the United Kingdom, family support at older ages is less affected by personal characteristics (such as parental divorce) than by factors reflecting need (e.g., poor health; Glaser et al., 2008). This also appears to be the case in transfers from parents to children: Older parents are more likely to provide help to adult children who are divorced and in poor health (Künemund, Motel-Klingebiel, & Kohli, 2005; Suitor, Pillemer, & Sechrist, 2006). Our results show that although mobility problems due to health were less prevalent in 2001/2 than in 1988/9, the impact of poor health on late-life support has largely remained unchanged; and once again, despite their better health, more fathers and mothers in early old age are giving and receiving help from children. Once older parents are in need of assistance, it appears that the family does step in to provide help regardless of other parental characteristics.

However, a number of factors need to be taken into account when considering the future. First, the size of the population needing care is likely to increase given projected increases in the number of older people with functional limitations (Pickard, Wittenberg, Comas-Herrera, King, & Malley, 2007). Second, although it is expected that spouses will continue to be the main source of support, they, too, are aging. Third, increasing survival among those with relatively heavy support needs (such as those with dementia) is likely to mean that some people will be too frail (physically and/or mentally) for family care and will need very intensive support (MacDonald & Cooper, 2007). Although this study's findings show that family support continues to be central, there remain concerns for the welfare of very old people in the future.

	Acknowledgments

 
This research is part of a project funded by the UK Economic and Social Research Council called "Disruptions in Family and Work Life: Implications for Support in Later Life" (RES-000-23-0662). The project also benefited from a visiting professor fellowship awarded by the University "La Sapienza," Rome, Italy, to Karen Glaser in June 2005. We are grateful to the UK Data Archive for permission and access to the British Household Panel Survey and the Survey of Retirement and Retirement Plans. We alone bear responsibility for the analysis and interpretation of the data.

K. Glaser planned the study, supervised the data analysis, ran the models, and wrote the paper. C. Tomassini helped plan the study, contributed to revising the paper, and devised the statistical analyses. R. Stuchbury was responsible for setting up and cleaning both data sets (including the creation of measures from the partnership and parenthood histories); she also contributed to revising the paper.

	Footnotes

 
Decision Editor: Kenneth F. Ferraro, PhD

Received for publication April 29, 2008. Accepted for publication July 4, 2008.

	References

TOP Abstract Methods Results Discussion References

 

• Ali, R., Greer, J., Matthews, D., Murray, L., Robinson, S., Sattar, G. (2006). General household survey 2006: Overview report. Retrieved July 4, 2008, from www.statistics.gov.uk/downloads/theme_compendia/GHS06/GHS2006overview.pdf.
• Allison, P. D. (1999). Comparing logit and probit coefficients across groups. Sociological Methods & Research, 28,186-208.[Abstract/Free Full Text]
• Aquilino, W. S. (1994). Impact of childhood family disruption on young adults' relationships with parents. Journal of Marriage and Family, 56,295-313.
• Askham, J., Ferring, D., Lamura, G. (2007). Personal relationships in later life. In J. Bond, S. Peace, F. Dittmann-Kohli, & G. J. Westerhof (Eds.), Ageing in society (pp. 186–208). London: Sage.
• Barrera, M. (1986). Distinctions between social support concepts, measures, and models. American Journal of Community Psychology, 14,413-445.
• Berthoud, R., Gershuny, J., (Eds.). (2000). Seven years in the lives of British families. Bristol, England: Policy Press.
• Bone, M., Gregory, J., Gill, B., Lader, D. (1992). Retirement and retirement plans. London: Her Majesty's Stationery Office.
• Bulcroft, K. A., Bulcroft, R. A. (1991). The timing of divorce, effects on parent-child relationships in later life. Research on Aging, 13,226-243.[Abstract]
• Cherlin, A. (1992). Marriage, divorce and remarriage. Cambridge, MA: Harvard University Press. (Original work published 1981).
• Cooney, T. M., Uhlenberg, P. (1990). The role of divorce in men's relations with their adult children after mid-life. Journal of Marriage and Family, 52,677-688.
• Disney, R., Grundy, E., Johnson, P. (1997). The dynamics of retirement: Analyses of the retirement surveys. London: Her Majesty's Stationery Office.
• Furstenberg, F. F., Hoffman, S. D., Shrestha, L. (1995). The effect of divorce on intergenerational transfers: New evidence. Demography, 32,319-333.[Medline]
• Ganong, L., Coleman, M. (2006). Obligations to stepparents acquired in later life: Relationship quality and acuity of needs. Journal of Gerontology: Social Sciences, 61B,S80-S88.[Abstract/Free Full Text]
• Gerstel, N. (1987). Divorce and stigma. Social Problems, 34,172-186.
• Glaser, K., Stuchbury, R., Tomassini, C., Askham, J. (2008). The long-term consequences of partnership disruption for support in later life in the United Kingdom. Ageing & Society, 28,329-351.
• Ha, J., Carr, D., Utz, R. L., Nesse, R. (2006). Older adults' perceptions of intergenerational support after widowhood: How do men and women differ? Journal of Family Issues, 27,3-30.[Abstract/Free Full Text]
• Haskey, J. (1993). First marriage, divorce, and remarriage: Birth cohort analyses. Population Trends, 72,24-33.
• Hermalin, A. I., (Ed.). (2002). The well-being of the elderly in Asia: A four-country comparative study. Ann Arbor: University of Michigan Press.
• House, J. S., Landis, K. R., Umberson, D. (1988, July 29). Social relationships and health. Science, 241,540-545.
• Kalmijn, M. (2007). Gender differences in the effects of divorce, widowhood, and remarriage on intergenerational support: Does marriage protect fathers? Social Forces, 85,1079-1104.
• Künemund, H., Motel-Klingebiel, A., Kohli, M. (2005). Do intergenerational transfers from elderly parents increase social inequality among their middle-aged children? Evidence from the German Aging Survey. Journal of Gerontology: Social Sciences, 60B,S30-S36.[Abstract/Free Full Text]
• MacDonald, A., Cooper, B. (2007). Long-term care and dementia services: An impending crisis. Age and Ageing, 36,16-22.[Abstract/Free Full Text]
• Malpass, P. (2008). Housing and the new welfare state: Wobbly pillar or cornerstone? Housing Studies, 23,1-19.
• Payne, G. (2006). Re-counting "illiteracy": Literacy skills in the sociology of social inequality. British Journal of Sociology, 57,219-240.[Medline]
• Pezzin, L. E., Schone, B. S. (1999). Parental marital disruption and intergenerational transfers: An analysis of lone elderly parents and their children. Demography, 36,287-297.[Medline]
• Pickard, L., Wittenberg, R., Comas-Herrera, A., King, D., Malley, J. (2007). Care by spouses, care by children: Projections of informal care for older people in England to 2031. Social Policy & Society, 6,353-366.
• Schooler, C., Revell, A. J., Caplan, L. J. (2007). Parental practices and willingness to ask for children's help later in life. Journal of Gerontology: Psychological Sciences, 62B,P165-P170.[Abstract/Free Full Text]
• Shapiro, A. (2003). Later-life divorce and parent-adult child contact and proximity. Journal of Family Issues, 24,264-285.[Abstract]
• Sigle-Rushton, W., Hobcraft, J., Kiernan, K. (2005). Parental divorce and subsequent disadvantage: A cross-cohort comparison. Demography, 42,427-446.[Medline]
• Soldo, B. J., Hill, M. S. (1993). Intergenerational transfers: Economic, demographic, and social perspectives. In G. G. Maddox & M. P. Lawton (Eds.), Annual review of gerontology and geriatrics (Vol. 13, pp. 187–216). New York: Springer.
• Stevenson, B., Wolfers, J. (2007). Marriage and divorce: Changes and their driving forces. Journal of Economic Perspectives, 21,27-52.[Medline]
• Stuchbury, R., Glaser, K., Tomassini, C., Askham, J. (2005, June/July). Married or single: Which shall I tick? Findings from a study of BHPS marital status data. Paper presented at the 2005 BHPS Research Conference, University of Essex, Colchester, Essex, England. Available online at www.iser.essex.ac.uk/bhps/2005/docs/pdf/papers/stuchbury.pdf.
• Suitor, J., Pillemer, K., Sechrist, J. (2006). Within-family differences in mothers' support to adult children. Journal of Gerontology: Social Sciences, 61B,S10-S17.[Abstract/Free Full Text]
• Taylor, A. (1994). Appendix: Sample characteristics, attrition and weighting. In N. Buck (Ed.), Changing households: The British Household Panel Survey 1990–1992 (pp. 291–311). Colchester, Essex, England: University of Essex, Economic and Social Research Council Research Centre on Micro-Social Change.
• Taylor, M. F. E., Brice, J., Buck, N., Prentice-Lane, E. (2007). British Household Panel Survey user manual volume A: Introduction, technical report and appendices. Retrieved July 4, 2008, from www.iser.essex.ac.uk/ulsc/bhps/doc/pdf_versions/volumes/bhpsvola.pdf.
• Thornton, A., Young-DeMarco, L. (2001). Four decades of trends in attitudes toward family issues in the United States. Journal of Marriage and Family, 63,1009-1037.
• Tomassini, C., Wolf, D. A. (2000). Stability and change in the living arrangements of older Italian women: 1990-1995. Genus, LVI,203-219.
• University of Essex, Institute for Social and Economic Research. (2006, June). British Household Panel Survey; Waves 1–14, 1991–2005 [Computer file; 2nd ed.]. Distributed by UK Data Archive. SN: 5151.
• Wachter, K. W. (1997). Kinship resources for the elderly. Philosophical Transactions of the Royal Society of London: Biological Sciences, 352,1811-1817.

This Article Abstract Full Text (PDF) Services Download to citation manager