Continuity and Change in Mothers’ Favoritism Toward Offspring in Adulthood

Abstract

The importance of parental favoritism in childhood and adulthood has been well documented; little is known, however, about changes over time in such within-family differentiation. Drawing on theories of life course processes and developmental psychology, the authors used 7-year panel data collected from 406 older mothers about their relationships with 1,514 adult children to explore patterns of favoritism regarding caregiving and emotional closeness. The findings demonstrated continuity in patterns of mothers’ favoritism. Mothers tended to prefer the same children across time, particularly regarding preferred caregivers. It was anticipated that children’s social-structural characteristics, similarity to their mothers, structural position in the family, and support provision to mothers would predict favored child status across time; however, only similarity and support processes were strong and consistent predictors of change and continuity in patterns of mothers’ favoritism.

Despite a powerful norm of equal treatment of offspring, research over the last two decades has demonstrated convincingly that parents often differentiate among their children in such domains as closeness and support (Suitor, Sechrist, Plikuhn, Pardo, & Pillemer, 2008). Although much of this work has been the province of developmental psychologists, the role of within-family differences in parenting has also been of substantial interest to both sociologists (Conley, 2004; Steelman, Powell, Werum, & Carter, 2002) and economists (Becker, 1991), focusing on the way in which structural factors such as birth order and gender differentially affect the experiences and opportunities of children within the same family.

There is evidence that such patterns continue into adulthood. Early studies by Bedford (1992) and Baker and Daniels (1990) revealed that a substantial proportion of adults felt that their parents favored some children in the family over others, whereas both Aldous, Klaus, and Klein (1985) and Brackbill, Kitch, and Noffsinger (1988) found that most parents reported that they differentiated among their children in adulthood in terms of affection, pride, and disappointment. More recent studies have confirmed that differentiation is common as offspring reach middle age and parents move into their later years (Suitor et al., 2008; Suitor & Pillemer, 2013).

Research has shown that such within-family differentiation has important consequences for both adult children’s psychological well-being and for the quality of sibling relations in adulthood. Specifically, the perception that their parents favor one child over another in adulthood is associated with higher depressive symptoms among adult children (Pillemer, Suitor, Pardo, & Henderson, 2010) as well as greater conflict and less closeness among siblings (Boll, Ferring, & Filipp, 2005; Gilligan, Suitor, Kim, & Pillemer, 2013; Suitor et al., 2009; Suitor, Gilligan, Johnson, & Pillemer, 2013). Patterns of favoritism also shape the experiences of older mothers when they require assistance. Recent evidence has shown that when older mothers’ earlier preferences for particular adult child caregivers are violated when they require assistance several years later, they experience a decline in psychological well-being (Suitor, Gilligan, & Pillemer, 2013). Taken together, this body of research shows that parental favoritism is a salient factor in adult children’s well-being and sibling relations in adulthood as well as in mothers’ well-being when they require care.

Despite the growing interest in parental favoritism in middle age and beyond, almost all studies have been cross-sectional; therefore, little is known about how parental preference may change over time. In the only published longitudinal research on this topic, Boll, Michels, Ferring, and Filipp (2010) examined continuity of adult children’s perceptions of differential treatment by parents over a relatively short time period (2 years). Their findings suggest that differential treatment is generally a stable characteristic of relationships rather than subject to extensive change over short periods of time. No longitudinal study, however, has investigated changes in actual parental preference or over a longer period. A prospective longitudinal study is necessary to understand what factors may predict changes in parental preference over time.

In this study, we used data from 406 older mothers gathered at two points, 7 years apart, regarding their relationships with each of their 1,514 adult children. Drawing on life course and adult development theories, we posed two questions about continuity and change in maternal favoritism in adulthood: First, to what extent do patterns of mothers’ favoritism regarding emotional closeness and preferences for caregiving change across time? Second, what characteristics of mothers, adult children, and mother-child dyads shape which offspring remain favored or move between favored and unfavored status across time?

Continuity and Change in Parent-Adult Child Relations

The life course perspective draws on sociological theories of social change and its relationship to individual and family development (Elder, 1985, 1994; Moen & Hernandez, 2009; Settersten, 2006). Life course scholars emphasize that family members’ lives are inextricably linked; therefore, life events experienced by one individual affect the lives and relationships of others in the family (Moen & Hernandez). The dynamic dimension of this conceptualization makes it especially well suited to studying changes in family relations over the life course. In particular, life course theories suggest that alterations in the relationship between parents and adult children would be most likely to follow life events or status transitions experienced by either role partner. Previous research has shown that adult children’s behaviors and characteristics, rather than parents’ characteristics or life events, shape parental favoritism (Suitor et al., 2008); however, these findings have been based on cross-sectional designs that could not take into consideration either how mothers’ characteristics at any single point could affect changes in favoritism or how changes in mothers’ characteristics could affect these processes. Therefore, in this article we focus on changes in the lives of children while also taking into consideration characteristics of mothers that may help to explain changes in favoritism.

Explaining Patterns of Change in Maternal Favoritism

Adult Children’s Characteristics and Behaviors

The child-level factors we propose as predictors of changes in favoritism are based on four key dimensions of the life course perspective: (a) children’s social structural positions, (b) children’s structural positions in the family, (c) similarity between mothers and children, and (d) support processes.

Role of children’s social structural factors

We propose that children’s social structural characteristics will play an important role in maternal preferences; however, there are theoretical and empirical bases for alternative hypotheses regarding these effects. The first argument is that mothers’ preferences will be positively related to entry into or maintenance of normative adult statuses and negatively related to failure to achieve such statuses. A powerful social norm holds that children should attain adult statuses in a timely fashion, establish independent lives, and ultimately become potential sources of support for parents (Pillemer, Suitor, Mock, Sabir, & Sechrist, 2007; Ryff, Schmutte, & Lee, 1996). To the extent that adult children experience these normative transitions, relationship quality tends to improve (Suitor, Gilligan, & Pillemer, 2013). This pattern occurs because such transitions confirm that the adult child is conforming to societal norms regarding maturational development. When children do not achieve normative adult statuses, parents feel obligated to assist them, and such children serve as a reminder that parents have not achieved their task of socialization (Greenfield & Marks, 2006; Ryff et al., 1996). Thus, normative transitions—such as marrying, becoming a parent, and being employed—could be expected to increase the likelihood that a child would either continue to be favored or move into favored status, whereas problematic transitions (e.g., divorcing or losing one’s job) could be expected to increase the likelihood that a child would remain or move out of favored status.

Alternatively, theoretical and empirical work on the “greedy family” can be used to argue that children’s normative transitions into marriage, parenthood, and employment have the potential to limit the time and attention focused on relations with parents (Coser & Coser, 1974; Sarkisian & Gerstel, 2008), resulting in less close intergenerational ties. Thus, it is possible that such transitions, or the maintenance of such adult statuses, would make it more likely that children would remain or move out of being favored.

Taken together, these arguments do not provide the basis for a single hypothesis; therefore, we instead explored which of these perspectives best explains patterns of continuity and change in favoritism across time.

Role of structural positions in the family

Two structural positions in the family are also likely to play salient roles in mothers’ favoritism across time: (a) children’s deviant behaviors in adulthood and (b) birth order.

Adult children’s problems have been found to have a strong impact on parents’ psychological well-being (Green, Ensminger, Robertson, & Juon, 2006; Greenfield & Marks, 2006; Pillemer & Suitor, 1991); thus, it is not surprising that studies have found that such problems are associated with more negative relationships with parents (Birditt, Fingerman, & Zarit, 2010; Fingerman, Cheng, Birditt, & Zarit, 2011; Greenfield & Marks). The effects of children’s problems occur not only because such difficulties often place additional responsibilities on parents (Condry, 2007; Green et al.) but also because they cause parents to question their parenting skills (Green et al.; Ryff et al., 1996) and may cause embarrassment and shame (Condry; Green et al.). Such feelings are particularly likely when the problems involve illegal behaviors (Condry; Green et al.).

On the basis of this set of findings, we hypothesized that mothers would favor children who had never engaged in deviant behaviors, or who disengaged in such behaviors between waves and would not favor children who engaged in deviant behaviors at both time points or began engaging in deviant behaviors between waves.

Birth order may also play a role in patterns of favoritism across time. Although birth order has received little attention in the study of later-life families, research on within-family differences has revealed that mothers report being most emotionally close to last-borns (Suitor & Pillemer, 2007). We anticipated that the same pattern would be found in the present study, thus fostering stability in mothers’ choices across time. Previous analyses have found no effects of birth order on preferences for care, so we did not develop specific hypotheses regarding this relational context.

Similarity between mothers and adult children

Perceptions of similarity have been recognized as highly salient to parent-adult child relations since its designation as one of the core components of Bengtson and colleagues’ model of intergenerational solidarity (Bengtson, 2001; Silverstein & Bengtson, 1997). Cross-sectional studies have found perceived similarity to be one of the best predictors of the quality of parent-adult child relations using both between- and within-family designs (Pillemer, Munsch, Fuller-Rowell, Riffin, & Suitor, 2012; Rossi & Rossi, 1990; Suitor & Pillemer, 2013).

Few studies, however, have explored how changes in perceived similarity affect parent-child relations across time. Suitor’s (1987) year-long panel study of adult daughters’ return to college revealed that changes in daughters’ values and interests resulting from their enrollment increased closeness to well-educated mothers while decreasing closeness and increasing conflict with less educated mothers. Furthermore, a 10-year follow-up to the original study indicated that changes in the daughters’ perceived similarity to their mothers continued to shape their relationships long after the initial transition (Plikuhn, Suitor, & Powers, 2009).

On the basis of this evidence, we hypothesized that changes in mothers’ perceptions of similarity to their adult children would predict patterns of favoritism between Time 1 (T1) and Time 2 (T2). Specifically, we predicted that mothers would continue to prefer children whom they perceived as remaining or becoming more similar to them across time and not prefer children whom they viewed as remaining or becoming less similar to them.

Gender similarity is also likely to play an important role in mothers’ favoritism across time, despite its invariant nature. Child gender has been found to be one of the most consistent predictors of both affective relations and support exchange between parents and adult children (Suitor, Sechrist, Gilligan, & Pillemer, 2011). In particular, the mother-daughter tie has generally been found to be the closest, most enduring, and mutually supportive of all parent–child gender combinations. Furthermore, mothers often report being closer to daughters specifically because they share the same gender (Suitor & Pillemer, 2006). On this basis, we anticipated that, across both relational domains as well as across time, daughters would be favored over sons. Thus, gender, unlike most of the other factors we considered, is likely to promote stability rather than fuel change between waves.

Social support processes

Studies of intergenerational relations have often highlighted the association between parent-adult child exchange and relationship quality. This line of work has explored extensively the ways in which relationship quality shapes adult children’s patterns of support to their parents (Fingerman et al., 2011; Silverstein & Bengtson, 1997; Swartz, 2009; Van Gaalen & Dykstra, 2006). Less attention has been directed toward understanding the role that children’s support may play in parents’ feelings toward their offspring when parents are in their later years and are at greater risk of facing challenges for which they will need assistance. We propose that children’s support to their mothers increases the likelihood that those children are preferred over others in the family who do not provide such support. Furthermore, we propose that children who have a history of providing support to their mothers will be more likely to be preferred over those who only recently provided support. In this argument we draw from Carstensen’s theory of socioemotional selectivity (Carstensen, 1992; Carstensen, Fung, & Charles, 2003), which suggests that children who were a source of support when young would provide a greater sense of security and predictability later in life, leading their mothers to prefer them as their future caregivers and to feel more emotionally close to them than to their other offspring.

On the basis of this argument, we hypothesized that mothers would continue to prefer children with whom they had a history of support as well as children who had begun to provide support between T1 and T2 and would not prefer children from whom they no longer received support or from whom they did not report receiving support at either time point. We expected that instrumental and expressive support would play somewhat different roles in predicting mothers’ favoritism, with expressive support being more salient for emotional closeness and instrumental support being more salient for preferred caregivers.

Mother-Level Characteristics

Although mother-level characteristics have not been found to shape patterns of favoritism in cross-sectional studies (Suitor, Sechrist, & Pillemer, 2007), it is possible that changes in mothers’ characteristics may affect patterns of favoritism across time. In particular, changes in mothers’ age, health status, and marital status might produce changes in mothers’ likelihood of favoring some children over others.

On the basis of socioemotional selectivity theory (Carstensen, 1992; Carstensen et al., 2003), one could anticipate that as mothers age they would be more likely to favor specific children regarding both emotional closeness and preferences for caregiving. Carstensen argued that as people age and their time perspective alters, they focus on interpersonal relationships that are the most rewarding and increasingly withdraw from those that are not. Therefore, we hypothesized that the older mothers in the study would be more likely to favor some children over others than would their somewhat younger counterparts.

A similar argument can be made regarding experiencing declines in health, leading mothers to be more likely to specify particular adult children whom they would prefer, especially as caregivers. As has been shown in recent work on parental favoritism and caregiving, mothers are highly likely to prefer specific children following a major health event, and the violation of these preferences leads to higher depressive symptoms (Suitor, Gilligan, & Pillemer, 2013). This pattern occurs because the mothers select offspring on the basis of socioemotional characteristics that ensure predictability and harmonious relations—a pattern consistent with principles of socioemotional selectivity theory (Carstensen, 1992). Thus, we anticipated that mothers who had experienced declines in health would become more likely to favor some children over others than would mothers who had not experienced such events.

Third, we anticipated that changes in mothers’ marital status would shape patterns of favoritism. Because widowhood, rather than divorce, is the most common marital status change at this stage in the life course, we focused on this transition. Although mothers’ marital status generally plays no role in favoritism in later-life families (Suitor et al., 2007), mothers who are recently widowed may be less likely to differentiate among their offspring. This may occur because there is typically a notable increase in both closeness to and support from adult children following this transition (Fuller-Thompson, 2000; Khodyakov & Carr, 2009), thus making choices less necessary. Thus, we hypothesized that mothers who had become widowed between waves would be less likely to favor some children over others.

Family size may also play a role in mothers’ differentiation. When mothers have a larger number of children to choose among, continuity of favoritism may be lower than when there are fewer offspring.

Finally, residential proximity has been found to affect patterns of differentiation in later-life families (Suitor et al., 2011). We hypothesized that mothers would name children who remained or became more proximate to them as both preferred caregivers and as those to whom they were most emotionally close and would become less likely to name those who moved further away between waves.

Method

The data used in the present analyses were collected as part of the Within-Family Differences Study (WFDS). The design of the WFDS involved selecting a sample of mothers 65 through 75 years of age with at least two living adult children and collecting data from mothers regarding each of their children. (For a more detailed description of the WFDS design, see Suitor & Pillemer, 2006, where portions of this section have been published previously.) The first wave of interviews in the WFDS took place with 566 women between 2001 and 2003; the original study was expanded to include a second wave of data collection from 2008 through 2011. In this research we used data collected from 406 mothers who were interviewed at both T1 and T2 regarding 1,514 of their children.

Procedure

Massachusetts city and town lists were used as the source of the original WFDS sample. With the assistance of the Center for Survey Research at the University of Massachusetts, Boston, the first and third authors drew a probability sample of women ages 65 through 75 with two or more children from the greater Boston area. The T1 sample consisted of 566 mothers, who represented 61% of those eligible for participation, a rate comparable to that of similar surveys in the past decade (Wright & Marsden, 2010).

For the follow-up study, the survey team attempted to contact each mother who participated in the original study. At T2, 420 mothers were interviewed. Of the 146 mothers who participated at only T1, 78 had died between waves, 19 were too ill to be interviewed, 33 refused, and 16 could not be reached. Thus, the 420 represent 86% of mothers who were living at T2. Comparison of the T1 and T2 samples revealed that the respondents differed on subjective health, educational attainment, marital status, and race. Mothers who were not interviewed at T2 were less healthy, less educated, and less likely to have been married at T1; they were also more likely to be Black. There were no significant differences in their likelihood of having differentiated among their children at T1; in fact, consistent with all of the patterns thus far in the study, mothers’ likelihood of differentiation was driven by children’s rather than mothers’ characteristics. Specifically, mothers were less likely to differentiate if all of their children shared their values or were all daughters. This is not surprising, given that these are the two primary characteristics on which mothers differentiated within the family regarding emotional closeness and caregiving preferences (Suitor et al., 2007); thus, when all children shared these characteristics, mothers had little basis for differentiating. Comparisons between the mothers alive at T2 who did and did not participate revealed that they differed only on education and subjective health.

We omitted six mothers from the present analysis because they lost one of their two children between waves; we also omitted four mothers who were missing data on one or more of the dependent variables. Finally, for each of the two favoritism domains (caregiving and emotional closeness), we included data only from mothers who favored a child at either T1 or T2. Four mothers did not differentiate among their adult children across either of the relational contexts at either T1 or T2 and were therefore also omitted from the present analyses. The final analytic sample consisted of 406 mothers, each of whom differentiated at T1 only, T2 only, or both T1 and T2. The 406 mothers had a total of 1,514 living adult children at T2. Because the percentage of mothers who favored a child varied across domains, the number of children included in each analysis varied. For the analysis of caregiving preferences, the sample included 1,373 children; for the analysis of emotional closeness, the sample included 1,269 children. Listwise deletion was used to handle missing data on the independent variables because there were fewer than 1% missing on any variable in the analysis (cf. Allison, 2010).

The demographic characteristics of the 406 mothers and 1,514 children who were named as favorite at T1 or T2 for at least one of the relational domains are presented in Table 1.

Table 1.

Demographic Characteristics of Mothers and Adult Children

Mothers	M (SD) or % (n = 406)
Age in years (SD)	77.9 (3.2)
Race (%)
Black	25.1
Not Black	74.9
Marital status (%)
Married	38.9
Cohabiting	0.7
Divorced/separated	14.5
Widowed	45.3
Never married	0.5
Education (%)
Less than high school	19.1
High school graduate	44.1
Some college	12.9
College graduate	24.0
Number of children (SD)	3.8 (1.8)
Adult childrena
Age in years (SD)	49.5 (5.8)
Daughters (%)	51.0
Education (%)
Less than high school	7.0
High school graduate	32.2
Some college	12.5
College graduate	48.4

^an = 1,514.

Measures

Dependent variables

To determine maternal preference, at both T1 and T2 mothers were asked a series of questions that required them to select among their adult children. Each mother was asked to select which child (a) she would prefer help from if she (the mother) became ill or disabled and (b) to whom she felt the most emotionally close. Each child was coded as 0 for each of the items for which he or she was not chosen and as 1 for each item for which he or she was chosen. In cases in which mothers were initially unwilling to differentiate among their children, the interviewers were instructed to prompt the mothers with a follow-up question (e.g., “But is there one child whom you would talk to first?”). Fewer than 5% of the mothers were moved by the prompt to select a child, and there were no sociodemographic differences between mothers who did and did not respond to the prompt.

We created changes-in-favoritism measures for both emotional closeness and caregiving preferences by categorizing each child on the basis of whether the mother named him or her in each of the relational domains at four points: (a) neither T1 nor T2, (b) at T1 only, (c) at T2 only, or (d) at both T1 and T2. Dummy variables were created to reflect each of these categories. The distributions of the favoritism variables for emotional closeness and caregiving preferences are shown in Table 2.

Table 2.

Distribution of Children Favored at Time 1 (T1) and Time 2 (T2; in Percentages)

Time point when chosen	Care preferences (n = 1,395)	Most emotionally close (n = 1,290)
Not chosen at T1 or T2	64.2	64.3
Chosen at T1 only	9.7	11.7
Chosen at T2 only	12.0	14.6
Chosen at T1 and T2	14.1	9.5

Child- and dyad-level independent variables

Perceived value similarity was measured at both T1 and T2 by asking mothers “Which of your children would you say in general is the most similar to you?” Examination of the reasons mothers provided to explain their choices indicated that they chose on the basis of similarity of values, interests, life experiences, and approaches to life, consistent with Bengtson and colleagues’ conceptualization of normative solidarity (Bengtson, 2001; Silverstein & Bengtson, 1997). Each child was coded as 0 if she or he was not named as most similar and as 1 if she or he was named as most similar to the mother. Changes in perceptions of similarity measures were created by assigning each child to one of the following four categories: (a) as the most similar to her at neither T1 nor T2 (63.5%), (b) at T1 only (9.9%), (c) at T2 only (11.1%), or (d) at both T1 and T2 (15.5%). Dummy variables were created to reflect each of these categories.

To assess support processes, at both waves mothers were asked about the support they received from each of their adult children. For expressive support, mothers were asked, “In the past year, has [child’s name] given you: a) comfort during a personal crisis; or b) advice?” Each item was coded 0 or 1. We began by combining the two items into one measure of expressive support at each wave. We then created a set of dummy variables to reflect whether the child had provided expressive support to the mother at T1 only, T2 only, or at both T1 and T2. The distribution for expressive support provided by children was did not provide support at T1 or T2 (15.7%), (a) provided support at T1 only (10.6%), (c) provided at T2 only (19.3%), and (d) provided at both T1 and T2 (54.3%).

To assess instrumental support, mothers were asked whether, in the past year, the child had provided (a) help with household chores and (b) help when ill. We then created a measure of changes in the provision of instrumental support using the same procedures as used to create the measure of changes in expressive support. The distribution for instrumental support provided by children was (a) did not provide support at T1 or T2 (29.1%), (b) provided support at T1 only (16.2%), (c) provided at T2 only (17.0%), and (d) provided at both T1 and T2 (37.7%).

To measure children’s deviant behaviors, mothers were asked whether each of their adult children had experienced any of a series of problems. For the present analysis, we used substance abuse or problems with the law. At T1, the mothers were asked to specify whether the child had experienced these problems at any point in adulthood; at T2, they were asked whether the child had experienced these problems in the previous 5 years. Children were then assigned to one of the three following categories: (a) never engaged in deviant behaviors in adulthood (86.6%); (b) engaged in deviant behaviors in adulthood prior to T1, but disengaged in those behaviors by T2 (6.3%); and (c) engaged in deviant behaviors at T1 and T2 or began engaging in these behaviors between T1 and T2 (7.1%). Ideally, we would have liked to have created a separate category for “beginning deviant behaviors between T1 and T2”; however, there were few cases that fit this criterion (n = 40; 2.6%), as would be expected given the average age of the adult children at T2 (49.5 years).

Child’s marital status was reported by mothers at T1 and T2. Marital status was transformed into married (1) or not married (0) at each time point. Each child was assigned to one of the following four categories: (a) unmarried at both T1 and T2 (27.3%), (b) married at T1 only (6.5%), (c) married at T2 only (9.4%), (d) or married at both T1 and T2 (56.9%). A few had divorced and remarried between waves; we coded these children as married at T1 and T2. Similarly, those who had been unmarried at T1 but married and divorced between waves were coded as unmarried at T1 and T2.

To measure children’s parental status, mothers reported how many offspring each child had living at T1; at T2, they were asked whether each child had given birth to or adopted children between waves. Given the average age of the adult children at T2 (49.5 years), it is not surprising that only a small number became parents for the first time between T1 and T2; thus, it was not possible to include becoming a parent between waves as a separate category. Each child was categorized as being a parent (1) or having no children at T1 (0). Seventy percent of the adult children were parents at T1.

We asked mothers about their children’s employment at T1, but not about their current employment status at T2. Eighty-three percent of the adult children were employed at T1 (no = 0, 1 = yes). Instead, we collected information on children’s recent unemployment, rather than employment changes from T1 to T2. Mothers were asked whether each child had “not had a job when he/she wanted to work” in the previous year (no = 0, 1 = yes); 5.5% were unemployed and seeking work within in the year prior to T2.

Residential proximity was measured in travel time by ground transportation. The seven categories were (a) same house, (b) same neighborhood, (c) less than 15 minutes away, (d) 15 to 30 minutes away, (e) 30 to 60 minutes away, (f) more than 1 hour but less than 2 hours, and (g) and 2 or more hours away. The proximity variable at both waves was transformed into living “two or more hours away,” “more than an hour but less than two hours,” and “less than one hour.” For the change variable, each child was assigned to one of the following three categories: (a) “child lived more than two hours away at T1 and T2” (19.8%), (b) “child-moved further away between T1 and T2” (8.3%), or (c) “child moved within or remained within two hours at T1 and T2” (71.8%). Although we would have preferred to have further refined the final category, there were too few cases in which adult children moved closer between T1 and T2 to make this feasible.

To measure birth order, each child was coded as first-, middle-, or last-born, based on age. Previous research has shown that last-borns are favored by their mothers regarding emotional closeness, whereas first-borns are not (Suitor & Pillemer, 2007); for this reason, we dichotomized the variable into last-born (1) and not last-born (0). Twenty-seven percent of the offspring were last-borns.

Gender was coded 0=son, 1=daughter.

Mother-level characteristics

Family size was measured using the number of living adult children in the family at T2 (M = 3.9, SD = 1.8). Mothers’ age was measured in years (M = 77.9, SD = 3.2). To create the measure of mothers’ recent widowhood, we combined information provided by the mothers at T1 and T2 regarding their marital status at the time of the interviews. We classified mothers who became widowed between T1 and T2 as “recently widowed.” Slightly less than 14% (13.8%) became widowed between waves. Mothers were classified as having experienced a decline in health if, at T2, (a) they reported a serious injury or illness for which they needed assistance at some point within 2 years or (b) or if they experienced a new chronic condition for which they reported that they needed help within the same period. More than half of the mothers (57.2%) met this criterion.

Analytic Plan

Throughout the multivariate analyses, the parent-child dyad, rather than the parent, was the unit of analysis. In other words, the 1,514 children who are the units of analysis are nested within the 406 families on whose reports the present analysis is based; thus, the observations are not independent. To take this factor into account, we used multilevel multinomial logistic regression. This technique is well suited to our research question, which asked “What changes in children’s characteristics and behaviors predict mothers’ choices at T1 and T2, at T1 only, at T2, or not at either T1 or T2?” Furthermore, multilevel multinomial techniques allowed us to consider the role of mothers’ characteristics, in addition to children’s characteristics. The type of multinomial analysis we used creates a set of estimates that is one less than the number of categories because each category is compared to the reference category (Heck, Thomas, & Tabata, 2012). Thus, for both dimensions of favoritism we had four nominal outcomes. For example, in the analyses of mothers’ preferred caregiver, the referent category was adult children who were not named by their mothers as their preferred caregivers at either time point. They were compared to children in the other three categories—named as preferred caregivers at T1 only, at T2 only, and at both T1 and T2. To allow us to compare children who were named at both T1 and T2 with those named at T1 only, we conducted a second set of analyses, using “chosen at T1 only” as the referent category. The analyses were conducted using SPSS 20.

Results

Describing Continuity and Change in Mothers’ Favoritism

We began the multivariate analyses by examining the variance explained by the mother-level characteristics. We ran an intercept-only model, which provided the variance components to calculate the interclass correlation coefficients (Heck et al., 2012). The intraclass correlation coefficients ranged from .01 to .02, indicating that the mother-level factors accounted for only 1% to 2% of the variance in mothers’ choices. Although the amount of mother-level variance is small, we conducted subsequent analyses to determine whether we could identify any particular mother-level characteristics that accounted for this explained variance. This set of analyses revealed that, of the mother-level characteristics, only recent widowhood and family size predicted patterns of favoritism. Therefore, in an effort to be parsimonious, we omitted all other mother-level characteristics from the final models presented in the article.

The patterns of stability and change in mothers’ favoritism toward particular children are shown in Table 2. As shown in the left-hand columns, there was substantial continuity in patterns of mothers’ preferences regarding caregiving. Nearly two thirds of the children (64.2%) were not chosen at either T1 or T2, and 14% were preferred at both T1 and T2. Less than 10% were preferred at T1 but not T2, and 12% were preferred at T2 but not T1.

As shown in the right-hand column of Table 2, there was also continuity in mothers’ reports of the children to whom they were most emotionally close at T1 and T2. Nearly two thirds of the offspring were not chosen at either time point (64.3%), and slightly less than 10% were chosen as most close at both T1 and T2. Just under 12% of children were reported as most emotionally close at T1 but not T2, and about 15% were named at T1 but were not named again at T2.

Taken together, these findings suggest a remarkably high level of continuity from T1 to T2. For preferred caregiving, less than one quarter of the children moved into or out of preferred status, and in the case of emotional closeness, only slightly more than one quarter changed statuses.

Although our focus is primarily on continuity and change in children’s favoritism status, it is it interesting to note the high level of continuity of mothers who named the same children across time. Of those mothers who named a child at both T1 and T2, 63.8% named the same children as their preferred caregivers at both waves, and 61.9% named the same children as those to whom they were most emotionally close at both waves (results not shown.)

Predicting Continuity and Changes in Mothers’ Favoritism

Next, we turn to the question of what factors led particular children to remain favored, remain unfavored, or move in and out of being favored between T1 and T2. Tables 3 and 4 both include three columns of odds ratios (ORs); the first shows the ORs of a child having been chosen at T2 only, and the second shows the ORs of having been chosen at both T1 and T2. In these two columns, comparisons are to children who were not named as the mother’s choice at either wave. In contrast, the third column presents the ORs of having been chosen at both T1 and T2 relative to having been chosen at T1 only. Thus, this column indicates which children remained favored across the study, as opposed to those who were favored at T1 but were no longer favored at T2. We use the phrase “preferred at both T1 and T2” to distinguish children preferred at both waves from those preferred at only T1 or only T2.

Table 3.

Multilevel Multinomial Logistic Regression Predicting Mother’s Choice for Caregiving (n = 1,373)

Predictor	Child chosen at T2 only (ref. = not chosen)	Child chosen at both T1 and T2 (ref. = not chosen)	Child chosen at both T1 and T2 (ref. = chosen at T1)
Mother-level characteristics
Family size	0.76**	0.75**	0.93
Mother widowed between T1 and T2a	0.51*	1.01	1.65
Children’s social structural characteristics
Child marital statusb
Ended marriage between T1 and T2	1.42	0.99	0.75
Entered marriage between T1 and T2	0.65	0.66	0.89
Remained married from T1 to T2	0.85	0.81	0.84
Child is parent	0.73	0.86	1.05
Employmentc
Employed at T1	0.82	0.69	0.82
Looking for work year before T2	0.59	0.88	0.49
Children’s structural positions in the family
Child engaged in deviant behaviors in adulthoodd
Reported T1 only	1.57	0.80	1.13
Reported T2 only or both T1 and T2	0.74	0.57	0.84
Last-borne	1.14	1.59*	1.50
Child provided support to mother
Provided instrumental supportf
T1 only	1.88	1.54	0.69
T2 only	2.88**	0.92	0.65
T1 and T2	3.13**	2.63**	0.91
Provided expressive supportg
T1 only	1.47	1.49	1.76
T2 only	1.13	2.23*	2.95*
T1 and T2	1.53	2.54*	3.15**
Similarity between mother and child
Child most similar to motherh
T1 only	1.88*	2.82**	0.90
T2 only	2.04**	2.82**	1.11
T1 and T2	3.28**	6.92**	2.05*
Child is daughter	6.42**	19.07**	3.64**
Residential proximity to motheri
Moved further between T1 and T2	1.27	1.61	1.92
Lived within 2 hours at T1 and T2 or moved within 2 hours by T2	1.40	1.52	1.58
Log likelihood		2,208.83
AIC		2,360.92
BIC		2,729.02

Note: T2 = Time 2; T1 = Time 1; ref. = referent; AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion.

^aReferent is not recently widowed.

^bReferent is child unmarried at T1 and T2.

^cReferent is child employed at T1.

^dReferent is child did not engage in deviant behaviors in adulthood.

^eReferent is not last-born.

^fReferent is child did not provide instrumental support T1 or T2.

^gReferent is child did not provide expressive support T1 or T2.

^hReferent is child not most similar at T1 or T2.

ⁱReferent is child lived 2+ hours away at T1 and T2.

^*p ≤ .05.

^**p < .01.

Table 4.

Multilevel Multinomial Logistic Regression Predicting Mother’s Choice for Emotional Closeness (n = 1,269)

Predictor	Child chosen at T2 only (ref. = not chosen)	Child chosen at both T1 and T2 (ref. = not chosen)	Child chosen at both T1 and T2 (ref. = chosen at T1)
Mother-level characteristics
Family size	0.81**	0.76**	0.84*
Mother widowed between T1 and T2a	0.84	0.93	0.96
Children’s social structural characteristics
Child marital statusb
Ended marriage between T1 and T2	0.67	0.60	0.77
Entered marriage between T1 and T2	0.44*	0.65	1.18
Remained married from T1 to T2	0.96	0.52*	0.71
Child is parent	0.75	0.84	1.16
Employmentc
Employed at T1	1.12	0.73	0.92
Looking for work year before T2	1.41	1.84	1.06
Child’s structural positions in the family
Child engaged in deviant behaviors in adulthoodd
Reported T1 only	1.93*	0.58	1.09
Reported T2 only or both T1 and T2	0.97	0.52	0.75
Last-borne	1.50*	2.90**	1.59
Child provided support to mother
Provided instrumental supportf
T1 only	1.11	1.09	1.18
T2 only	1.28	0.85	0.73
T1 and T2	1.82*	1.42	0.81
Provided expressive supportg
T1 only	0.95	0.77	0.84
T2 only	0.92	1.17	0.77
T1 and T2	1.07	1.10	1.23
Similarity between mother and child
Child most similar to motherh
T1 only	1.66	2.64**	1.10
T2 only	2.21**	3.69**	2.33*
T1 and T2	2.72**	7.62**	2.05*
Child is daughter	2.22**	1.83**	1.04
Residential proximity to motheri
Moved farther between T1 and T2	0.81	0.60	1.37
Lived within 2 hours at T1 and T2 or moved within 2 hours by T2	1.03	0.88	1.10
Log likelihood		2,260.10
AIC		2,412.88
BIC		2,774.61

Note: T2 = Time 2; T1 = Time 1; ref. = referent; AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion.

^aReferent is not recently widowed.

^bReferent is child unmarried at T1 and T2.

^cReferent is child employed at T1.

^dReferent is child did not engage in deviant behaviors in adulthood.

^eReferent is not last-born.

^fReferent is child did not provide instrumental support T1 or T2.

^gReferent is child did not provide expressive support T1 or T2.

^hReferent is child not most similar at T1 or T2.

ⁱReferent is child lived 2+ hours away at T1 and T2.

^*p ≤ .05.

^**p < .01.

Preferences for Care

The findings for mothers’ preferences for care are presented in Table 3. The first two rows show the ORs for the effects of mother-level characteristics on mothers’ choice of any particular child as their preferred caregivers. As shown in the first row, the ORs for preferring any particular child as caregiver were notably smaller as family size increased at T2 (OR = 0.76) and at both T1 and T2 (OR = 0.75), compared to never being chosen. As shown in the second row of T able 3, the ORs of preferring any particular child were substantially smaller at T2 only when mothers became widowed between waves (OR = 0.51).

We now turn to the question of which characteristics and behaviors led particular children to remain preferred caregivers, remain not preferred caregivers, or move in and out of being preferred between T1 and T2. Contrary to expectations, as shown in the next several rows of Table 3, mothers’ caregiving preferences were not predicted by children’s marital status, parental status, employment status, or deviant behaviors. Birth order was also a less consistent predictor than anticipated, playing a role only in predicting which children mothers preferred at both T1 and T2 (OR = 1.59) compared to children who were never chosen. Specifically, mothers were more likely to prefer last-borns over other children in the family.

We hypothesized that both instrumental and expressive support would play roles in predicting preferences for caregiving. Consistent with this hypothesis, mothers were more likely to prefer children at T2 only who had provided them with instrumental support at T2 only (OR = 2.88) compared to children who were never chosen and to those who had never provided support. Also consistent with our hypothesis, mothers were more likely to prefer children at T2 only and at both T1 and T2 who had provided them with support at both T1 and T2 compared to those who were never chosen and to those who had never provided support (ORs = 3.13 and 2.63).

Provision of expressive support also predicted which children mothers were more likely to prefer, but only for those chosen at both T1 and T2. Specifically, mothers were more likely to prefer children at both T1 and T2 who had provided expressive support at T2 only (OR = 2.23), compared to those who were never chosen and who had never provided support. Mothers were also more likely to choose children at both T1 and T2 who provided support at T2 only (OR = 2.95) compared to those who were chosen at T1 only and had not provided support. A similar pattern can be seen for children who provided expressive support at both T1 and T2 (ORs = 2.54 and 3.15).

The next set of rows of Table 3 display the outcomes for similarity. Most striking about this set of findings is that, for all conditions in which the reference category is “never preferred as caregiver” (columns 1 and 2), mothers preferred care from the children whom they perceived as most similar to them relative to those whom the mothers never named as most similar. Furthermore, the OR was particularly high for naming children at both T1 and T2 whom the mothers reported as being most similar to them at both waves (OR = 6.92) compared to those who had never been preferred and never been most similar to them. Also, as shown in the right-hand column, children who were perceived as most similar to their mothers at both T1 and T2 were more likely to continue to be preferred as caregivers across the period of the study compared to those who were never most similar and who were preferred at T1 only (OR = 2.05).

Gender was the strongest predictor of mothers’ preferences for caregivers, with ORs ranging from 19.07 for being chosen at both T1 and T2 (compared to never being chosen), to 3.64 for being chosen at both T1 and T2 (compared to T1).

Finally, residential proximity did not predict mothers’ choice of caregivers at any time point.

Emotional Closeness

The findings regarding children to whom mothers were most emotionally close are presented in Table 4. The first two rows show the ORs for the effects of mother-level characteristics on mothers’ reports of being most emotionally close to any particular child. As shown in the first row, the ORs of naming any particular child as most emotionally close were notably smaller as family size increased for all comparisons. As shown in the second row, mothers’ recent widowhood played no role in predicting mothers’ favoritism regarding emotional closeness.

Next, we consider which characteristics and behaviors led particular children to remain chosen as most emotionally close to the mothers, remain unchosen, or move in and out of being chosen between T1 and T2. Children’s marital status played only a small role regarding to which children mothers felt the greatest emotional closeness, predicting closeness only under two conditions. First, mothers were notably less likely to favor children at T2 only who became married between waves (OR = 0.44) compared to those who remained unmarried and were never chosen as most emotionally close. Second, mothers were less likely to report being most emotionally close to children at both T1 and T2 who remained married across the study (OR = 0.52) compared to those who remained unmarried and never chosen. Contrary to expectations, as shown in the next several rows of the table, emotional closeness was not predicted by children’s parental status or employment status.

Although children’s deviant behaviors predicted emotional closeness under only one condition, this finding demonstrates the importance of changes in serious norm violations. Mothers were more likely to name as most emotionally close at T2 those children who had engaged in deviant behaviors but who had ceased these behaviors by T2 (OR = 1.93) compared to those who were never chosen. In other words, offspring who had disengaged from deviant behaviors were chosen at T2 over those who had not engaged in deviant behaviors in adulthood at any point.

Birth order was a consistent predictor of mothers’ favoritism regarding emotional closeness. Mothers were more likely to report being most emotionally close to their youngest children at T2 only (OR = 1.50) and at both T1 and T2 (OR = 2.90) compared to those who were never chosen.

As shown in the next several rows of Table 4, children’s support to their mothers predicted favoritism regarding emotional closeness in only one case. Specifically, mothers were more likely to choose children at T2 only who had provided instrumental support at both T1 and T2 compared to those who had never provided support and were never chosen (OR = 1.82).

The next set of rows in Table 4 displays the outcomes for similarity. As was the case for preferred caregiver, mothers’ perceptions of similarity were strong predictors of favoritism regarding emotional closeness. In fact, in only two cases did similarity not predict emotional closeness, and in some cases the strength of the relationship between similarity and closeness was notable (OR = 7.62).

Gender also predicted mothers’ favoritism regarding emotional closeness at T2 only (OR = 2.22) and at both T1 and T2 (OR = 1.83) compared to never being chosen.

Finally, residential proximity did not predict mothers’ favoritism regarding emotional closeness at any time point.

Analyses Using T1 Categories of the Independent Variables as Referents

We also conducted separate analyses using the “T1 only” categories of the independent variables as the reference categories to allow comparisons between children who, for example, provided support at T1 only and those who provided support at T1 and T2. We conducted these analyses using the “T1 only” category for provision of support to mother, value similarity, parental status, marital status, deviant behaviors, employment at T1, and proximity.

The analyses using T1 as the reference revealed that only changes in support to mothers predicted continuity and change in caregiving preferences and emotional closeness (tables not shown). First, children who continued to provide instrumental support at both T1 and T2 were substantially more likely to be preferred at T2 only compared to those who provided support only at T1 and were never preferred (OR = 1.76, p < .05). Second, children who continued to provide instrumental and expressive support were more likely to be preferred at both T1 and T2 as future caregivers compared to those who provided support at T1 only and who were never chosen (instrumental: OR = 1.78, p < .05; expressive: OR = 2.10, p < .05). Last, children who provided instrumental support at both T1 and T2 were more likely to be chosen as most emotionally close at T2 only compared to children who provided support only at T1 and were never chosen (OR = 1.78, p < .05).

Because value similarity is associated with mothers’ favoritism, mothers’ provision of support, and children’s structural characteristics, we conducted analyses using a nested approach to test for possible mediation effects and found that none of the other variables were mediated by value similarity.

Finally, because gender played such a prominent role throughout the analyses, we examined whether the patterns of findings differed by child’s gender. We ran the models separately for sons and daughters and found that the findings did not differ systematically from the full models (tables not shown). We also ran the models separately by birth order and found that the findings did not differ systematically from the full models (tables not shown).

Summary of Findings

The multivariate analyses indicated two clear patterns. First, perceived similarity played a central role in mothers’ favoritism across both dimensions of favoritism at both time points and across time. Second, child’s gender was a consistent predictor of mothers’ favoritism, with mothers preferring daughters for both caregiving and emotional closeness at both time points as well as across time. Third, mothers tended to favor their youngest children, as we anticipated.

Several characteristics of children and mother-child dyads did not predict favoritism as expected, however. For example, instrumental support predicted mothers’ preferences for caregiving in most but not all cases. Furthermore, contrary to our hypotheses, expressive support predicted preferences for care but did not predict favoritism regarding emotional closeness. Also, with the exception of marital status, none of the social structural factors showed any trend toward being predictors of favoritism, including children’s parental or employment status, and the analyses revealed no consistent patterns for residential proximity or children’s deviant behaviors.

Discussion

This article addressed the following question: Does maternal preference for particular children remain stable over time, such that the same offspring remain favored? To address this issue, we drew on the life course perspective, which emphasizes the linked lives of family members in different generations as well as the need to examine dynamic changes in older parent-adult child relationships as families move through time (Silverstein, 2005). Both theory and research based on the life course perspective suggested that parental preference would be a dynamic process, subject to change over a 7-year period.

Beyond identifying the degree to which parental favoritism changes in later-life families, it is important to examine why such changes occur. Following the life-course perspective’s emphasis on studying both continuity and change in family relations, we proposed that transitions in adult children’s social structural positions would affect whether they were or were not preferred across time. We extended our conceptual framework beyond social structural factors, in response to the need identified by Mayer (2009) to integrate developmental and social components of the life course within individual studies. We used Carstensen’s developmental theory of socioemotional selectivity to expand the life course perspective (Carstensen, 1992; Carstensen et al., 2003), hypothesizing that parental preference would gravitate toward adult children who have more rewarding characteristics (value similarity, history of support) and fewer unrewarding characteristics (deviant behaviors).

Continuity in Patterns of Favoritism

On the basis of the life course perspective, we anticipated that substantial change would occur in mothers’ favoritism over the 7-year study interval, given the dynamic nature of middle-aged adult children’s lives and social selectivity processes. For both dimensions of favoritism, however, we found considerable continuity in both the actual patterns of favoritism and in the factors that predict those patterns. For both preferences for care and emotional closeness, approximately three quarters of the adult children continued to be named at both T1 and T2 or to not be named at either wave. such continuity is relatively high in the context of findings from other studies. In fact, studies of changes in support networks typically report lower levels of stability, even across much shorter periods of time (Morgan, Neal, & Carder, 1997; Suitor & Keeton, 1997; Suitor & Pillemer, 1996; Wellman, Wong, Tindall, & Nazer, 1997). Thus, in the context of similar studies, the level of stability we have found is notable.

Predictors of Favoritism Over Time

We found that patterns of favoritism across time were predicted by a combination of continuity and change in the independent variables. We hypothesized that when mothers’ perceptions of their adult children’s value similarity changed, mothers’ preferences would change as well. Consistent with this expectation, when mothers perceived children as becoming more similar to them, those offspring were more likely to become chosen at T2 for both caregiving and emotional closeness. Continuity, however, also played an important role in these processes: Mothers were substantially more likely to prefer children who remained similar to them at both T1 and T2 compared to children who had never been perceived as similar. Furthermore, the likelihood of preferring the same children across both waves was greatest when offspring remained as those whom mothers perceived as most similar. A similar trend was seen in children’s provision of instrumental and expressive support to their mothers, although the findings were not as consistent as for similarity.

In some cases, neither continuity nor change predicted patterns of maternal favoritism. In particular, children’s employment and parental status did not predict favoritism at either T1 or T2. Such normative achievements by children when they are entering adulthood may be of greater importance to mothers as they attempt to successfully launch their children. At the point at which children reach middle age, however, mothers are entering a stage in the life course when they are more concerned with structuring their own lives in a way that emphasizes continuity and harmony and avoids interpersonal stress (Carstensen, 1992; Carstensen et al., 2003). This pattern may help to explain why the structural changes in children’s lives found to be salient in earlier stages of the life course (Aquilino, 1997; Kaufman & Uhlenberg, 1998) played such a small role in patterns of favoritism among the mothers in the present study.

In contrast to other social structural positions, mothers’ favoritism was in some cases influenced by their children’s marital status. Mothers were less likely to be most emotionally close at T2 to offspring who became married between waves and to those who remained married at both T1 to T2. This pattern may reflect the perspective that marriage is often a “greedy institution” (Coser & Coser, 1974; Sarkisian & Gerstel, 2008) in that the marital relationship is expected to take precedence over other kin relations. These findings provide support for the hypothesis that marriage may have detrimental consequences on parent-child relations despite the fact marriage is a normative achievement.

In one case—children’s deviant behaviors— only change predicted alternations in mothers’ favoritism. Specifically, offspring who had experienced problems with the law or with substance abuse in adulthood but who no longer engaged in these deviant behaviors by T2 were more likely to be named as those to whom the mothers were most emotionally close. This pattern is especially interesting because these children were preferred over others who had never engaged in these behaviors, suggesting that the mothers were relieved by the improved behaviors.

This finding is consistent with a considerable body of research demonstrating that problems in the lives of adult children have a significant negative effect on their parents (Greenfield & Marks, 2006; Milkie, Beirman, & Schieman, 2008; Pillemer & Suitor, 1991). The effects are particularly pronounced for deviant behaviors such as incarceration (Green et al., 2006) and substance abuse problems (Oreo & Ozgul, 2007). To our knowledge, however, no study has examined the effects of adult children’s transitioning out of deviant behaviors on either the parent-child relationship or parent outcomes. Thus, this area is particularly promising for future investigation.

Although in most cases mothers’ characteristics played only a small role in shaping patterns of favoritism across time, mothers were less likely to favor any of their offspring at T2 when they had become widowed between waves. This pattern reflects the finding of other studies that mothers become closer to their children in general following widowhood and are provided with support from a larger proportion of offspring at this time (Fuller-Thompson, 2000; Khodyakov & Carr, 2009), thus providing fewer bases for mothers to differentiate.

Taken together, the results we have presented indicate that mothers’ choices are strongly influenced by continuity and change in children’s characteristics and behaviors. This pattern is consistent with theories of adult development that emphasize increased selection of predictability and positive experiences in interpersonal relations in later life. In particular, the findings reflect Carstensen and colleagues’ socioemotional selectivity framework (Carstensen, 1992; Carstensen et al., 2003), in which they argue that individuals in late middle age and beyond place priority on relationships that are most rewarding.

It is worth noting that the most consistent predictor of mothers’ favoritism at any one time and across time was child gender. This finding would be expected, given the greater closeness between mothers and daughters across the life course as well as in patterns of family caregiving (Suitor et al., 2011). Thus, although gender was not a major theoretical focus of the study, it is important to highlight the salience of this characteristic in explaining continuity in mothers’ favoritism.

The consistency of the predictors across both dimensions of favoritism raises the question of whether mothers are naming the same children in both domains. This is not the case. As we have reported elsewhere (Suitor, Gilligan, & Pillemer, 2010), at T1 only 41% of the children named for one relational domain were named for both, and at T2 only 47% were named for both.

Limitations and Directions for Future Research

The study points toward several directions for future research. First, our analyses considered only mothers’ favoritism. Given well-established gender differences in parent-adult child relations (Suitor et al., 2011), it is possible that patterns of change and predictors of favoritism may differ for fathers. In fact, recent research contrasting mothers’ and fathers’ differentiation among their adult children has shown that fathers’ favoritism has more negative effects on sibling relations than does mothers’ (Gilligan et al., 2013). Thus, understanding stability and change in fathers’ favoritism is worthy of consideration.

Second, we examined patterns of change in maternal preference from the perspective of older parents rather than their children; previous research has shown that adult children and parents report differences in patterns of favoritism in the family (Suitor, Sechrist, Steinhour, & Pillemer, 2006). Thus, it is likely that adult children’s perceptions of patterns of their parents’ favoritism across time might vary considerably from those of their parents. Given that both children’s and parents’ perceptions of favoritism affect children’s depressive symptoms (Pillemer et al., 2010), studying patterns of favoritism across time from the children’s perspectives would increase our understanding of the role of within- family differences in offspring’s well-being.

Future research should continue to examine patterns of continuity and change in mothers’ preference for children, particularly regarding caregiving. Our findings revealed that there is limited movement in mothers’ choices for a caregiver. In fact, mothers’ preferences in the “pre-caregiving” years (65–75) were largely the same 7 years later, when care needs were more common. Recent research has shown that violation of these preferences has detrimental consequences for mothers’ psychological well-being (Suitor, Gilligan, & Pillemer, 2013). Thus, it is important to understand the processes by which particular children become caregivers and the role of parental favoritism in these processes.