Can Positive Social Exchanges Buffer the Detrimental Effects of Negative Social Exchanges? Age and Gender Differences
Katherine L. Fiori a Tim D. Windsor b Elissa L. Pearson c Dimity A. Crisp d
a Gordon F. Derner Institute of Advanced Psychological Studies, Adelphi University, Garden City, N.Y. , USA;
b School of Psychology, Flinders University, Adelaide , c
School of Psychology, The University of South Australia, Adelaide , and d
Centre for Research on Ageing, Health and Wellbeing, The Australian National University, Canberra , Australia
tion, partner status, and physical functioning. Results: We found that positive exchanges buffered against negative ex-changes for younger old adults, but not for older old adults, and for women, but not for men. Conclusions: Our findings are interpreted in light of research on individual differences in coping responses and interpersonal goals among late middle-aged and older adults. Our findings are in line with gerontological theories (e.g. socioemotional selectivity the-ory), and imply that an intervention aimed at using positive social exchanges as a means of coping with negative social exchanges might be more successful among particular pop-ulations (i.e. women, ‘younger’ old adults).
Supportive social relationships have long been of inter-
est to gerontologists given their capacity to contribute to both physical and psychological health in late middle-aged and older adults [e.g. 1–3 ]. It is widely recognized that social support can ameliorate stress and facilitate coping with transitions (e.g. retirement), in addition to contributing directly to well-being in the absence of stressful events [4] . However, the fact that interpersonal relationships involve costs as well as benefits should not be overlooked [5, 6] . Negative social exchanges are par-
Key Words
Social relations � Social support � Mental health � Socioemotional selectivity theory
Abstract
Background: Findings from existing research exploring whether positive social exchanges can help to offset (or‘buffer’ against) the harmful effects of negative social ex-changes on mental health have been inconsistent. This could be because the existing research is characterized by differ-ent approaches to studying various contexts of ‘cross-do-main’ and ‘within-domain’ buffering, and/or because the na-ture of buffering effects varies according to sociodemo-graphic characteristics that underlie different aspects of social network structure and function. Objective: The pur-pose of this study was to examine whether the buffering ef-fects of global perceptions of positive exchanges on the link between global negative exchanges and mental health var-ied as a function of age and gender. Method: We used a se-ries of regressions in a sample of 556 Australian older adults (ages 55–94) to test for three-way interactions among gen-der, positive social exchanges, and negative social exchang-es, as well as age and positive and negative social exchanges, in predicting mental health, controlling for years of educa-
Received: January 10, 2012 Accepted: May 29, 2012 Published online: July 18, 2012
Katherine L. Fiori, PhD Gordon F. Derner Institute of Advanced Psychological StudiesAdelphi University, 158 Cambridge Avenue Garden City, NY 11530 (USA) Tel. +1 516 877 4809, E-Mail fiori @ adelphi.edu
ticularly noxious stressors, exerting substantial and det-rimental effects on psychological [5, 6] and physical health [7] . It may be especially important to understand how positive aspects of social relationships can offset in-terpersonal stressors in an older adult population, for whom continued exposure to stressors can have substan-tial health effects given a heightened vulnerability to dis-ease and to immune-compromising stress [7] .
The purpose of the present study was to add to and clarify the growing body of literature that considers how positive aspects of social relationships can ameliorate the detrimental mental health effects of negative aspects of social relationships in late middle-aged and older adults. Specifically, findings from existing research exploring whether positive social exchanges can help to offset (or ‘buffer’ against) the harmful effects of negative social ex-changes on mental health have been inconsistent [4, 6, 8–10] . One possible explanation for these inconsistent findings is that the nature of buffering effects varies ac-cording to sociodemographic characteristics that under-lie different aspects of social network structure and func-tion. Thus, our primary aim was to disentangle whether buffering effects vary as a function of age or gender, which are two sociodemographic variables that have doc-umented associations with a variety of social network characteristics [11–14] .
Definitions and Measurement Issues
Exchanges with network members (i.e. spouse, part-ner, family members, friends, coworkers) can be both positive (supportive) and negative (stressful). Although positive social exchanges are commonly understood and studied as consisting of multiple dimensions (e.g. infor-mation, instrumental, and emotional support) [11] , nega-tive social exchanges have typically not been studied within this multidimensional framework [5] . However, research suggests that both positive and negative social exchanges are multidimensional [5] , and that when com-paring positive and negative social exchanges it is impor-tant to have symmetry in terms of the measures (e.g. di-mensions, intensity) and in the time frame assessed (e.g. over the past month) [12, 15] .
To overcome these issues, in the present study we rely on the Positive and Negative Social Exchange scale (PANSE) [16] , which considers multiple domains of positive and neg-ative exchanges that are designed to be parallel in nature and are measured over an equal time interval (in the past month). Specifically, we define ‘positive social exchanges’
as interactions with network members that involve the re-ceipt of four important domains of social support: infor-mational support, instrumental support, emotional sup-port, and companionship. Positive social exchanges, then, include situations in which a network member offers help-ful advice, does a favor, says kind things, or provides good company. ‘Negative social exchanges’ are assessed with four domains that are ‘parallel’ to the four domains of pos-itive social exchanges; namely, unwanted advice or intru-sion, failure to provide help, unsympathetic or insensitive behavior, and rejection or neglect. Negative social ex-changes can therefore be understood as potent forms of interpersonal stressors. Furthermore, the PANSE is de-signed to assess overall perceptions of network exchanges (i.e. participants are prompted to think in general about ‘the people in their lives,’ such as their spouses or partners, family members, friends, neighbors, in-laws, or others), al-lowing us to obtain a holistic impression of the extent to which broad positive and negative social exchanges inter-act to predict mental health [e.g. 17 ].
Positive and Negative Exchanges: The Buffering
Hypothesis
For decades, clear links have been drawn between pos-itive social exchanges (e.g. experience of care, support and companionship with network members) and im-proved psychological and physical health [1–3, 18–20] . Conversely, negative social exchanges (e.g. tension, con-flict, and/or neglect from network members) have been found to be potent stressors, exerting substantial and det-rimental effects on psychological [5, 6, 8, 21] and physical health [7, 22, 23] . Although it has long been recognized that positive social exchanges can buffer against the neg-ative impacts of a range of stressors, research in the area has typically focused on non-interpersonal sources of stress [24] . There are, however, reasons to believe that the availability of positive social exchanges could be particu-larly important in buffering against the detrimental ef-fects of negative social exchanges. Given that common negative events include negative social feedback and so-cial rejection, positive exchanges that facilitate favorable self-evaluations and cognitive reappraisal of problematic relationships may reduce the detrimental impact of nega-tive social exchanges. Positive exchanges can also mini-mize the stress response by increasing perceptions of available support [2, 25] . Positive social exchanges may facilitate emotion-focused coping by assisting the indi-vidual in cognitively reframing negative exchanges and
Positive and Negative Social Exchanges Gerontology 3
managing the associated negative emotions that may arise [26, 27] . Positive social exchanges may also boost self-esteem and reduce or change dysfunctional attitudes, which have been demonstrated in previous studies to help account for the relationship between negative social ex-changes and psychological distress [28] .
Empirical research concerned with the potential buff-ering effects of positive exchanges on the relationship be-tween negative exchanges and mental health have pro-duced mixed findings, ranging from no buffering [6] , to support for the buffering hypothesis [8–10] , to reverse buffering, whereby social support is found to amplify the harmful effects of negative exchanges [4, 10] . To under-stand these mixed findings, it is important to distinguish among ‘cross-domain’ buffering (i.e. when positive ex-changes in one relationship or in a particular category of relationships buffers against the adverse effects of nega-tive exchanges in a different relationship or category of relationships), ‘within-domain’ buffering (i.e. when buff-ering occurs within the same relationship or within the same category of relationships), and more ‘global’ buffer-ing (as in the present study, when individuals are askedto consider their network as a whole).
The inconsistent findings in the existing research are perhaps not surprising in light of the mixed use of cross-domain and within-domain buffering, as well as the vari-ations in types of domains examined. Some researchers have looked solely at one type of buffering. For example, Schuster et al. [8] examined only within-domain buffering in a large sample of married couples, and found evidence for the existence of buffering against depressed mood for both men and women, but only within the relationship domain ‘relatives’ (not friends or spouse). Other research-ers have compared cross-domain to within-domain buff-ering, but with a focus on different domains. For example, in a study of over 200 college students, researchers found evidence for cross-domain buffering but not within-do-main buffering when examining psychological distress over time; however, the study focused only on ‘friend’ and ‘roommate’ relationships [9] . In contrast, research exam-ining depressive symptoms in older adults (aged 60–92) and younger adults (aged 28–59) found primarily cross-domain buffering for the older adults and within-domain buffering for the younger adults (‘domains’ consisted of spouse, child, and other relatives/friends) [10] . The only study using more global assessments of the social network to examine buffering included a very small sample of adult caregivers, and uncovered the existence of reverse-buffer-ing [4] . Specifically, the researchers who conducted this study found that caregivers who rated their networks as
helpful had depression scores that increased steeply with increasing upset, whereas caregivers who rated their net-works as not very helpful had depression scores that were relatively unaffected by increasing upset.
One possible explanation for the inconsistent findings is that the existing research is characterized by different approaches to studying various contexts of ‘cross-do-main’ and ‘within-domain’ buffering. Within-domain measures of positive and negative exchanges tend to be correlated, which makes finding statistical interactions to examine buffering effects difficult [9] (and could ex-plain the null findings reported in several studies). An-other possible explanation for the inconsistent findings is that the nature of buffering effects varies according to sociodemographic characteristics related to social net-work structure and function. Age and gender are two such characteristics. Although some researchers have ex-amined men and women or younger and older adults sep-arately in looking for buffering effects [9, 10] , three-way interactions statistically testing for such differences have not been considered, and as mentioned above, results from these studies have been complicated by the focus on cross-domain versus within-domain associations. Our focus in the present study was on examining whether the buffering effects of global perceptions of positive ex-changes on the link between global negative exchanges and mental health varied as a function of age and gender.
Implicit to the buffering hypothesis described above is the notion that positive social exchanges with network members can ameliorate the experience of stress (includ-ing interpersonal stress). However, the complex nature of interpersonal development and motivation also points to possible differences among sub-groups in the extent to which positive exchanges might be expected to act as a buffer. Gender and age are two important characteristics that may influence the extent to which positive social ex-changes buffer the effects of negative social exchanges on psychological health, given established gender and age differences in social motivation and social network com-position [11–14] . In the present study, we aimed to clarify the nature of buffering effects involving global assess-ments of positive and negative social exchanges by focus-ing on age and gender differences.
Gender and the Buffering Effect
According to the ‘tend-and-befriend’ perspective on human biobehavioral responses to stress [29] , evolution-ary selection associated with differential parental invest-
Fiori /Windsor /Pearson /Crisp
Gerontology4
ment results in men and women typically displaying dif-ferent patterns of responses to stress. Specifically, the greater role of women in rearing young offspring (and their responsibility for protecting both themselves and their offspring when under threat) results in their devel-oping greater skills in, and a greater reliance on, making effective use of the social group when dealing with stress. In contrast, men may have evolved to deal with threaten-ing situations in a manner more in keeping with tradi-tional conceptions of the ‘fight-or-flight’ stress response.
The tendency among women to be more socially inte-grated relative to men is supported by findings of numer-ous studies indicating that women typically have larger, denser, and more diverse social networks [11] , and report more positive as well as more negative social exchanges [8] . Furthermore, women may be both more negatively affected by social negativity than men [8] , and may re-ceive greater benefits to well-being from positive ex-changes in times of stress [30] . Finally, women more fre-quently seek out positive social exchanges as a means of coping compared to men [31] . Taken together, the avail-able evidence (i.e. the more diverse social networks, greater vulnerability to negative exchanges, and greater reliance on positive exchanges as a method of coping among women) points to the possibility that the buffer-ing effects of positive social exchanges on the relationship between negative exchanges and mental health are likely to be more evident among women than men.
To date, relatively few studies have examined the pos-sible moderating role of gender on the buffering effect of positive social exchanges. However, there is some evi-dence that interpersonal stress buffering through posi-tive social exchanges may be more common among wom-en than men. Walen and Lachman, in a study of nearly 2,400 married adults aged 25–75, found evidence primar-ily for cross-domain buffering among women, but not men; namely, among women, supportive friends offset the negative impact of a strained partner relationship on life satisfaction and positive mood, and both family and friend support buffered the negative effects of family strain on well-being [32] . In contrast, as mentioned ear-lier, Schuster et al. [8] examined only within-domain buffering in over 1,500 married couples aged 18 through 65, and did not find evidence of differential buffering (i.e. buffering was found for both men and women within the relationship category ‘relatives’). Similarly, Okun and Keith [10] did not find evidence for gender moderation in their study of adults aged 28–92. However, the substantial correlations between within-domain sources of positive and negative exchanges in these studies make it difficult
to effectively evaluate and interpret the gender differenc-es (or lack thereof) in buffering [cf. 9 ]. In the present study, we examined possible gender differences in buffer-ing using global assessments of positive and negative so-cial exchanges.
Age and the Buffering Effect
Gerontological theories of emotion regulation and in-terpersonal behavior point to the buffering role of posi-tive exchanges varying as a function of age. According to socioemotional selectivity theory (SST) [33] , an increas-ing salience of limited future time that corresponds with advancing age results in a reprioritizing of social and emotional goals. Specifically, as individuals age, they tend to invest in smaller social networks comprised of those individuals with whom they share emotionally meaningful relationships. The aging-related changes outlined by SST suggest that the buffering role of positive exchanges might become less important for maintaining mental health at older ages, as older adults may be likely to avoid those network members with whom negative ex-changes provide a significant source of stress by ‘win-nowing’ them from their network. In contrast, younger-old adults might experience a broader range of unavoid-able negative social exchanges (e.g. from coworkers or older children living at home), resulting in the availabil-ity of positive exchanges as a potential buffer taking on greater importance. In fact, research shows that people become more positive and less negative about their social relationships with age [33, 34] , and that older adults re-port interpersonal tensions as less stressful than do younger and middle-aged adults [13] .
Research concerned with age differences in interper-sonal problem-solving also suggests that positive ex-changes may become less important as a buffer to nega-tive exchanges with advancing age. Blanchard-Fields [35] found that older adults are more effective interpersonal problem solvers than are younger adults, reporting a greater use of passive emotion regulation strategies such as avoidance or withdrawal from conflict relative to younger adults. Similarly, Sorkin and Rook [26] found that maintaining harmony with an interaction partner was the most commonly endorsed primary coping goal among an elderly sample (aged 65–90) of over 400 indi-viduals. An effective repertoire of interpersonal coping skills, as well as a capacity to effectively regulate emo-tional responses to interpersonal tensions, may make old-er adults less reliant on positive exchanges with network
Positive and Negative Social Exchanges Gerontology 5
members as a means of maintaining emotional well-be-ing in the context of interpersonal stress.
Few studies have examined the possible moderating role of age on the buffering effect of positive social ex-changes. Walen and Lachman [32] found that friend sup-port buffered the impact of friend strain on health but only among older (aged 60–75) and not younger (aged 25–59) adults. Okun and Keith [10] also reported age dif-ferences in buffering relationships between younger and older adults, but as mentioned earlier their results indi-cated age differences in the source of support acting as a buffer (e.g. within vs. across social domains) rather than age differences in the presence or absence of buffering ef-fects. Again, however, the moderate to large correlations between within-domain sources of positive and negative exchanges in these studies make it difficult to interpret the interactions [cf. 9 ]. Furthermore, from a theoretical perspective, it may be a more promising starting point to examine age differences in the buffering nature of posi-tive exchanges across the entire network rather than within or across specific domains (e.g. spouse, child), since the theoretical ‘winnowing’ that occurs with age is concentrated on peripheral network members, not close ties [33] .
The Present Study and Hypotheses
The purpose of the present study was to shed new light on the nature of the buffering effect of positive social ex-changes on the link between negative social exchanges and mental health. Our main aim was to examine the role of sociodemographic variables, namely gender and age, which are associated with both the availability and expe-rience of positive and negative social exchanges, and which might account for inconsistencies in the buffering literature. In order to reduce problems of limited statisti-cal power, as well as limited generalizability of results that can arise from focusing on buffering effects that pertain to specific network domains (e.g. buffering within and across domains of family, spouses, and friends), we used the PANSE [16] . As mentioned earlier, the PANSE also considers multiple domains of positive and negative ex-changes that are designed to be parallel in nature. Impor-tantly, although research shows that measuring these multiple dimensions is necessary to ensure content valid-ity, research also indicates that the dimensions are indica-tors of more abstract (latent) constructs (i.e. ‘positive so-cial exchanges’ and ‘negative social exchanges’) [5] . Therefore, in the present study, we combine the multiple
dimensions into these broader subscales (‘positive’ and ‘negative’ social exchanges).
Our three hypotheses were as follows. First, we hy-pothesized that positive social exchanges and negative social exchanges would interact in predicting mental health. Specifically, we predicted that positive social ex-changes would buffer against the negative effects of neg-ative social exchanges on mental health. Second, given the gender differences in the use of positive exchanges as a means of responding to stress outlined above, we ex-pected to find three-way interactions among gender and positive and negative social exchanges in the prediction of mental health, such that the ‘buffering effect’ of posi-tive exchanges would be stronger among women relative to men. Third, given gerontological theories of emotion regulation and relationships [33] , as well as the reported age differences in social exchanges summarized earlier in this paper, we predicted a three-way interaction among age and positive and negative social exchanges, such that the ‘buffering effect’ of positive exchanges would be less evident at older ages.
In addition to age and gender, analyses were conduct-ed controlling for education and partner status, since ed-ucation [36, 37] and marital status [38, 39] have implica-tions for both mental health and the kinds of social ex-changes in which individuals engage. We also controlled for physical functioning given its implications for social exchanges [40, 41] and associations with mental health [42, 43] .
Method
Participants and Procedure The sample consisted of 556 community-dwelling adults re-
cruited from the Australian Capital Territory as a comparison sample to be contrasted with older adults relocating from the community to a retirement village. Initially, 2,000 individuals aged 55 years and over were randomly selected from the Austra-lian electoral roll (voting is compulsory for Australian citizens aged over 18, with some rare exceptions). Twenty-seven individu-als were excluded because their address details indicated that they lived in a retirement village (community-dwelling participants were our target given the broader project aims). The remaining 1,973 individuals were mailed a questionnaire and letter inviting them to participate. A total of 561 participants returned the ques-tionnaire, a response rate of 28.4%. Although the response rate was low, comparisons with the Australian Bureau of Statistics 2006 census data on adults 55–94 years indicated that the age and gender distributions in the sample were representative of the gen-eral population [44, 45] . Furthermore, our response rate com-pares favorably with other Australian survey research that has used a similar sampling procedure [46] .
Fiori /Windsor /Pearson /Crisp
Gerontology6
Because 5 participants had missing values for age, our final sample size was 556. Missing values ranged from 0 to 6.3% on the variables used in the current analysis, and individuals with miss-ing data were excluded on an analysis-by-analysis basis. The final sample consisted of 269 men and 287 women, with a mean age of 65.38 (SD = 8.29). The majority of the sample was married (73%), with an average of 13.87 years of education (SD = 3.05; see table 1 ). Ethics approval was obtained from the Australian National Uni-versity Committee for Ethics in Human Research.
Measures Positive and Negative Social Exchanges Social exchanges were measured using the PANSE [16] . This
scale consists of 24 items, with 12 items representing four dimen-sions of positive exchange (informational support, instrumental support, emotional support, and companionship) and 12 items representing four parallel dimensions of negative exchange (un-wanted advice or intrusion, failure to provide help, unsympathet-ic or insensitive behavior, and rejection or neglect). Participants were asked to consider the people in their lives (partner or spouse, family members, friends, neighbors, in-laws, or others) and indi-cate on a 5-point scale from 0 (never) to 4 (very often) how often various exchanges had occurred over the past month. Total scores were obtained for positive and negative exchanges by calculating mean responses for each subscale, with total scores ranging from 0 to 4 and higher scores representing greater levels of positive( � = 0.90) or negative ( � = 0.90) social exchanges.
Mental Health and Physical Functioning Self-rated mental health was assessed using the mental health
component score of the RAND-12 [47] . This score consists of 6 items; two items assess whether participants have had problems with work or other regular daily activities (e.g. ‘accomplished less than you would like’) as a result of ‘any emotional problems (such as feeling depressed or anxious)’ in the past 4 weeks (yes/no); three items assess how participants have felt in the past 4 weeks on a scale from 1 (all of the time) to 6 (none of the time) (e.g. ‘Have you felt calm and peaceful?’), and one item assesses how much of the time (in the past 4 weeks) the participants’ physical health or emo-tional problems interfered with social activities (like visiting with
friends, relatives, etc.), on a scale from 1 (all of the time) to 5 (none of the time). Factor loadings were applied based on an analysis of US population-based data, which produced scale scores with a mean of 50 and a standard deviation of 10. Higher scores indicate better mental health ( � = 0.80).
Although mental health was our primary outcome variable of interest, physical functioning was included as a control variable. Physical functioning was assessed using the physical functioning subscale of the Short-Form 36 Health Survey [48] . This subscale consists of 10 items assessing how participants’ health limits them in various activities (e.g. ‘walking several blocks’), with responses on a scale from 1 (yes, limited a lot) to 3 (no, not limited at all)( � = 0.91). Responses are recoded to range from 0 to 100 and then averaged, with higher scores indicating better physical functioning.
Sociodemographic Characteristics In addition to age and gender, analyses were conducted con-
trolling for total years of education (calculated based on educa-tional history) and partner status, 0 (unpartnered) or 1 (part-nered).
Statistical Analysis To test our first hypothesis of the existence of an ‘overall’ buff-
ering effect, we conducted a two-step hierarchical multiple linear regression predicting mental health, in line with generally estab-lished procedures [49] . Step 1 included control variables (gender, age, education, partner status, and physical functioning), as well as positive and negative social exchanges. Step 2 included the two-way interaction between positive and negative social exchanges (created by mean centering the social exchange variables and cre-ating a cross-product term).
To test our second and third hypotheses (i.e. whether age and gender are moderators of the ‘buffering effect’ of positive social exchanges), it was necessary to test for the existence of significant three-way interactions. Thus, after including all associated lower- order two-way terms in step 2, we added three-way interaction terms at a third step in multiple linear regressions predicting mental health. At each step of the model, we examined the change in R 2 to determine if the addition of the interactions made a sig-nificant contribution to the model. If the change in R 2 was sig-
Table 1. Means and percentages, standard deviations, and intercorrelations among all study variables (n = 556)
R anges: age 55–94; years of education 0–18; positive and negative social exchanges 0–4; mental health is a standardized t score. Physical functioning ranges from 0 to 100.
† p < 0.10, * p < 0.05; ** p < 0.01; *** p < 0.001.
Positive and Negative Social Exchanges Gerontology 7
nificant, we then examined the significance of the individual in-teractions.
We used plots of predicted values to help interpret the nature of significant interactions. Specifically, we plotted negative ex-changes on the original continuous scale along the x-axis, for groups defined according to high and low levels of positive social exchanges (based on a median split) such that the ‘buffering ef-fect’ (or lack thereof) would be obvious. We created separate plots for men and women to illustrate any three-way interactions with gender. In order to illustrate any three-way interactions with age, we used a median split to distinguish between younger and older participants in the sample (i.e. 55–63, n = 278, and 64–94, n = 278).
Results
Descriptive Analysis Table 1 provides means, standard deviations, and cor-
relations among all study variables. As can be seen in the table, older adults were less well educated, less likely to be partnered, reported fewer negative social exchanges, and had worse physical functioning than younger adults in the sample. Women were less well educated, less like-ly to be partnered, reported more positive social ex-changes, and had worse physical functioning and mental health than men in the sample. Also interesting to note is that partnered individuals reported more positive so-cial exchanges and better physical functioning and men-tal health. In addition, positive social exchanges were as-sociated with greater mental health and marginally bet-ter physical functioning, whereas negative social exchanges were associated with worse mental health. In line with our assumption outlined earlier in the paper that global (i.e. across entire network) assessments of positive and negative social exchanges should be uncor-related, the correlation between positive and negative ex-changes was not significant, r = –0.05. Mental health and physical functioning were positively correlated, r = 0.44, p ! 0.001.
Primary Analyses Overall Buffering Effect Table 2 shows the hierarchical linear regression pre-
dicting mental health from age, gender, education, part-ner status, physical functioning, negative and positive so-cial exchanges, and the interaction between negative and positive social exchanges. Step 1 shows that negative ex-changes negatively ( � = –0.21, p ! 0.001) and positive ex-changes marginally positively ( � = 0.08, p = 0.053) pre-dicted mental health. The addition of the two-way inter-action between positive and negative social exchanges to the second step of the hierarchical regression did not ex-
plain significant additional variance in mental health ( � R 2 = 0.003, p = 0.133).
Gender Moderation Table 3 shows results of the hierarchical linear regres-
sion used to test interactions among negative and positive social exchanges and gender. The addition of the three-way interaction (gender ! negative exchanges ! posi-tive exchanges) and lower order terms explained signifi-cant additional variance ( � R 2 = 0.021, p ! 0.05). Figure 1 illustrates the significant three-way interaction ( � = 0.12, p = 0.05); as predicted, positive social exchanges acted as a buffer for women but not for men in the sample.
Age Moderation Table 4 shows results of the model with inclusion of the
three-way interaction of age ! negative exchanges ! positive exchanges. Inclusion of the three-way interaction (and lower order two-way terms) improved model fit ( � R 2 = 0.010, p ! 0.05), indicating that the buffering ef-fect of positive exchanges on the relationship between negative exchanges and mental health varied as a func-tion of age. Figure 2 illustrates the significant three-way interaction (age ! negative exchanges ! positive ex-changes, � = –0.18, p = 0.038) with a median split of age
Table 2. Hierarchical multiple linear regression predicting men - tal health from negative and positive social exchanges and their interaction
for ease of presentation. As seen in the figure, positive social exchanges acted as a buffer for ‘younger old’ adults (aged 55–63) but not for the ‘older old’ adults (aged 64–94) in the sample.
Discussion
Research shows that positive social exchanges can have benefits for psychological health, primarily by buffering against stress [25] . On the other hand, nega-tive social exchanges can act as particularly noxious stressors [16, 17] . Thus, the question arises as to wheth-
er positive social exchanges can ‘buffer’ against the tox-ic effects of negative social exchanges. Findings from re-search on this topic have been inconsistent [4, 6, 8, 10] , in part because sociodemographic factors that could moderate the buffering effect have not been thoroughly investigated. The findings from the present study sug-gest that positive social exchanges are more likely to buffer against the detrimental mental health effects of negative social exchanges for women (not men) and for ‘younger old’ (and not ‘older old’) adults. We examine and interpret our findings in more detail in the discus-sion below.
Table 3. Hierarchical multiple linear regression predicting men - tal health from negative and positive social exchanges and their interactions with gender
� R 2 from step 1 to step 2 = 0.016, p = 0.011; � R 2 from step 2 to step 3 = 0.005, p = 0.050.
Table 4. Hierarchical multiple linear regression predicting men - tal health from negative and positive social exchanges and their interactions with age
� R 2 from step 1 to step 2 = 0.004, p = 0.409; � R 2 from step 2 to step 3 = 0.006, p = 0.038.
Positive and Negative Social Exchanges Gerontology 9
Overall Buffering Effect Contrary to our first hypothesis, we did not find a sig-
nificant interaction between positive and negative social exchanges predicting mental health. We had expected a ‘buffering’ effect, in which for individuals with ‘high’ lev-els of positive social exchanges, the association between
negative social exchanges and mental health would be less pronounced. However, this null result is not altogeth-er surprising since we also hypothesized three-way inter-actions such that this buffering effect would only be pres-ent for certain subgroups of the population. The presence of three-way interactions alters any interpretations of
60
55
50
Men
tal h
ealt
h p
red
icte
d v
alue
45
400 0.50 1.00 1.50
Negative social exchanges
2.00 2.50 3.00
Positive socialexchanges
(median split)
Low; Low; Low: R2 linear = 0.480
High; High; High:R2 linear = 0.592
a
Low
High
Low
High
70
60
50
40
Men
tal h
ealt
h p
red
icte
d v
alue
30
200 0.50 1.00 1.50
Negative social exchanges
2.00 2.50 4.003.00 3.50
Positive socialexchanges
(median split)
Low; Low; Low: R2 linear = 0.540
High; High; High:R2 linear = 0.104
b
Low
High
Low
High
Fig. 1. Interaction between positive and negative social exchanges predicting mental health for males ( a ) and females ( b ).
60
50
40
Phys
ical
hea
lth
pre
dic
ted
val
ue
30
0 1.00
Negative social exchanges
2.00 4.003.00
Positive socialexchanges
(median split)
Low; Low; Low: R2 linear = 0.610
High; High; High:R2 linear = 0.012
a
Low
High
Low
High
55
45
40
Phys
ical
hea
lth
pre
dic
ted
val
ue
250 1.00
Negative social exchanges
1.50 3.002.00
Positive socialexchanges
(median split)
Low; Low; Low: R2 linear = 0.069
High; High; High:R2 linear = 0.124
0.50 2.50
30
35
50
b
Low
High
Low
High
Fig. 2. Interaction between positive and negative social exchanges predicting mental health for adults aged 55–63 ( a ) and adults aged 64–94 ( b ).
Fiori /Windsor /Pearson /Crisp
Gerontology10
two-way interactions [50] , and it is likely that the overall buffering effect is not significant because it only operates for a portion of the sample. We turn now to the three-way interactions that were uncovered.
Gender and the Buffering Effect As expected, among women, greater negative ex-
changes were associated with worse mental health; how-ever, this association was stronger in the context of low levels of positive exchanges. At high levels of positive ex-changes, negative exchanges and mental health were less closely associated. In contrast, for men, negative exchang-es were negatively (but comparatively more weakly) as-sociated with mental health regardless of the levels of pos-itive exchanges. Our results are consistent with the find-ings from Walen and Lachman’s study of married adults; specifically, they found that supportive friends buffered the negative effects of partner and family strain on psy-chological health, but in all instances only among women and not men [32] . They also found that family support buffered the negative effects of family strain among women only.
Whereas Walen and Lachman [32] looked at buffering within and across specific relationship types, our results point to the possibility that positive exchanges more gen-erally buffer against the negative effects of negative ex-changes for women but not men, across the entire social network. Research shows that women more frequently seek out social support (i.e. positive social exchanges) as a means of coping than do men [31, 32] , and that women’s relationships are more likely to depend on emotional closeness than men’s [51] . This implies that women may be more likely to use positive exchanges (with anyone in their network) as a resource to cope with (any) negative exchanges.
The findings are also consistent with the ‘tend-and-befriend’ theory, which posits that biobehavioral re-sponses to stress differ for men and women [29] . Specifi-cally, whereas the traditional ‘fight-or-flight’ response may characterize the stress responses of men, women may be better served in protecting themselves and their offspring by rallying social network resources in timesof stress.
Age and the Buffering Effect In the present study, we also found a significant three-
way interaction among age, positive exchanges, and neg-ative exchanges predicting mental health. The findings indicated that the buffering effect of positive exchanges on the association between negative exchanges and men-
tal health was more strongly evident among ‘younger old’ adults than it was among ‘older old’ adults, consistent with our predictions. Our results are in line with geron-tological theories of emotion regulation and relationships (e.g. SST [33] ), which suggest that older adults prefer more intimate and emotionally meaningful ties. Since main-taining harmonious relationships seems to be an increas-ingly important goal with age [26, 33] , it may be that old-er individuals are more likely to use emotion-focused coping [27] in reaction to interpersonal stressors (e.g. for-giveness), which are strategies that do not necessarily re-quire seeking out positive social exchanges with others. In other words, as adults age, positive social exchanges may be used more as a way to promote emotionally satis-fying relationships, and less as a way of dealing with neg-ative aspects of relationships (e.g. negative social ex-changes; and hence we would be less likely to see ‘buffer-ing’ in this population).
Furthermore, SST suggests (and research confirms) that people become more positive and less negative about their social relations with age [33, 34] . In fact, in the pres-ent sample we found that the ‘older old’ adults reported significantly fewer negative interactions than the ‘youn-ger old’ adults. It could be that for the ‘younger old’ adults in our sample, most of whom are likely still in the work-force, negative social exchanges might be less avoidable and come from a broader range of network members (e.g. from coworkers or cohabitant children), whereas for the ‘older old’ adults, negative exchanges might be more eas-ily avoided. This reasoning is also in line with SST [33] , which suggests that adults tend to discard the more pe-ripheral members of their social networks as they age. Thus, since interpersonal tensions are less frequent and less stressful [13] for older compared to younger adults, perhaps there exists less of a need for ‘buffering’.
Interestingly, our findings are in contrast to those of Walen and Lachman [32] , who found only one instance of a three-way interaction among age, support, and strain; namely, that (source-specific) friend support buffered friend strain when predicting health, but only for older adults (60–95), not younger (25–39) or middle-aged (40–59) adults. In contrast, Okun and Keith [10] found buffer-ing effects primarily for older adults (ages 60–92) when positive and negative social exchanges were associated with different sources, whereas the buffering effects found for younger adults (ages 28–59) were found pri-marily when negative and positive exchanges were asso-ciated with the same source. The present study provides some clarification of the existing buffering literature by providing evidence in support of the role of positive ex-
Positive and Negative Social Exchanges Gerontology 11
changes as a stress buffer against negative exchanges, particularly for women and young-old adults, when glob-al, balanced measures of perceived network exchange quality are considered.
Limitations and Future Research There are several limitations to this study that should
be kept in mind when interpreting the results. First, as just suggested, the study is cross-sectional and correla-tional; thus, we cannot be sure of causal direction. Our models were based on the theoretical assumption that positive and negative social exchanges influence mental health, but it could be that individuals with poor mental health are more likely to experience negative social ex-changes. However, this potential for ‘reverse causality’ would not necessarily explain the buffering effect we found; that is, if individuals with poorer mental health engaged in more negative social exchanges, we would have no theoretical reason to expect positive exchanges to buffer this effect. Furthermore, longitudinal research does not seem to show evidence of this ‘reverse causality’ [6] . However, more longitudinal research is needed, and the field would also benefit from creative experimental approaches aimed at determining directionality. It was also the case that while our participants were drawn from a random population sample, the response rate for the survey was low, which may limit generalizability of the results to the broader population.
Although the global measures of positive and negative social exchanges used in the present study provided par-simonious models and minimized complexity stemming from comparing source-specific to cross-domain buffer-ing, it should be acknowledged that research shows vari-ation (at least by age) in terms of whether the buffering is source specific or exists across domains [10, 32] . Further-more, interactions with various social partners may have differential effects on mental health; for example, nega-tivity with spouse and relatives seems to be more detri-mental to mental health than negativity with friends [10, 12] . In addition, the sources of negative exchanges may vary as a function of gender [12] . Paradoxically, however, because source-specific measures of support and strain tend to be more highly correlated than global measures, interactions between the two (and hence buffering ef-fects) may be more difficult to detect when using such measures [32] . Future research might take a different ap-proach towards examining the source-specific buffering nature of positive exchanges, perhaps with daily diary as-sessments that could capture the ‘ups’ and ‘downs’ with-in specific relationships at more of a ‘micro’ level.
Finally, although we postulated that the different use of coping styles may be one reason why we see greater buffering among women and younger old adults than among men and older old adults, we did not specifically examine coping in this study. Future research should ex-amine coping styles and other psychological constructs as possible mechanisms to explain the differential buffer-ing effect. For example, positive social exchanges may support emotion-focused coping [27] by helping the indi-vidual reframe or reappraise the negative social exchang-es [16] , or by helping to manage the negative emotions aroused by the negative exchange. Other research indi-cates that the link between negative social interactions and distress may be accounted for by self-esteem, dys-functional attitudes, and external control beliefs [28] ; positive social exchanges may ‘buffer’ these negative ef-fects, then, by boosting the individual’s self-esteem, by changing his or her attitude towards the negative interac-tion, or by refocusing the individual’s control beliefs to be more internal. If these mechanisms work differently for men and women, or across age groups, they could offer potential explanations for the differential buffering ef-fects found in the present study and in other research.
In particular, for understanding the age moderation it is important to keep in mind the age differences in phys-ical functioning (with older adults experiencing worse functioning). Although we controlled for physical func-tioning in all of our analyses, longitudinal follow-up as-sessments would be necessary to determine if perhaps the poorer health of older adults could result in more restrict-ed social engagement [40] , which in turn could result in older adults relying on coping methods other than posi-tive social exchanges (e.g. avoidance) for dealing with in-terpersonal stress.
Conclusions
Findings from research on the buffering effects of pos-itive social exchanges have been inconsistent [4, 6, 9, 10] , perhaps in part because sociodemographic differences have not been properly addressed, and because the re-search has primarily focused on ‘within-domain’ and ‘cross-domain’ buffering. The present study tackles this gap in the research by specifically examining age and gender differences in the interaction between positive and negative social exchanges predicting mental health, and by using global (but multidimensional and parallel) assessments of positive and negative exchanges. There are, of course, practical implications of our findings, spe-
Fiori /Windsor /Pearson /Crisp
Gerontology12
cifically in terms of potential support interventions. Co-hen [25] suggests that social integration and support in-terventions should (1) involve several different aspects of the social environment (e.g. reducing negative exchang-es), and (2) be catered towards individual differences (i.e. based on variability in the effectiveness of support due to participants’ social skills, social traits, etc.). Our findings
imply that an intervention aimed at using positive social exchanges as a means of coping with negative social ex-changes might be more successful among particular pop-ulations (i.e. women, ‘younger old’ adults). Future re-search aimed at uncovering mechanisms can offer fur-ther clarity in developing effective social support interventions.
References
1 Antonucci TC, Akiyama H: An examination of sex differences in social support among older men and women. Sex Roles 1987; 17: 737–749.
2 Cohen S, Wills TA: Stress, social support, and the buffering hypothesis. Psychol Bull 1985; 98: 310–357.
3 House JS, Landis KR, Umberson D: Social re-lationships and health. Science 1988; 241: 540–545.
4 Pagel M, Erdly W, Becker J: Social networks: we get by with (and in spite of) a little help from our friends. J Pers Soc Psychol 1987; 53: 793–804.
5 Finch J, Okun M, Pool G, Ruehlman L: A comparison of the influence of conflictual and supportive social interactions on psy-chological distress. J Pers 1999; 67: 581–621.
6 Rook KS: Promoting social bonding: strate-gies for helping the lonely and socially iso-lated. Am Psychol 1984; 39: 1389–1407.
7 Newsom JT, Mahan TL, Rook KS, Krause N: Stable negative social exchanges and health. Health Psychol 2008; 27: 78–86.
9 Lepore SJ: Social conflict, social support, and psychological distress: evidence of cross-do-main buffering effects. J Pers Soc Psychol 1992; 63: 857–867.
10 Okun MA, Keith VM: Effects of positive and negative social exchanges with various sources on depressive symptoms in younger and older adults. J Gerontol B Psychol Sci Soc Sci 1998; 53: 4–20.
11 Antonucci TC, Jackson J: Social support, in-terpersonal efficacy, and health: a life course perspective; in Carstensen LL, Edelstein BA (eds): Handbook of Clinical Gerontology. New York, Pergamon, 1987, pp 291–311.
12 Bertera E: Mental health in U.S. adults: the role of positive social support and social neg-ativity in personal relationships. J Soc Pers Relat 2005; 22: 33–48.
13 Birditt KS, Fingerman KL, Almeida DM: Age differences in exposure and reactions to interpersonal tensions: a daily diary study. Psychol Aging 2005; 20: 330–340.
14 Fiori KL, Smith J, Antonucci TC: Social net-work types among older adults: a multidi-mensional approach. J Gerontol B Psychol Sci Soc Sci 2007; 62: 322–330.
15 Rook KS: Positive and negative social ex-changes: weighing their effects in later life. J Gerontol B Psychol Sci Soc Sci 1997; 52: 167–169.
16 Newsom JT, Rook KS, Nishishiba M, Sorkin DH, Mahan TL: Understanding the relative importance of positive and negative social exchanges: examining specific domains and appraisals. J Gerontol B Psychol Sci Soc Sci 2005; 60: 304–309.
17 Rook KS: Parallels in the study of social sup-port and social strain. J Soc Clin Psychol 1990; 9: 118–132.
18 Orth-Gomer K, Wamala SP, Horsten M, Schenck-Gustafsson K, Schneiderman N, Mittleman MA: Marital stress worsens prog-nosis in women with coronary heart disease: The Stockholm Female Coronary RiskStudy. JAMA 2000; 284: 3008–3014.
19 Uchino BN, Cacioppo JT, Kiecolt-Glaser JK: The relationship between social support and physiological processes: a review with em-phasis on underlying mechanisms and im-plications for health. Psychol Bull 1996; 119: 488–531.
20 Seeman TE, Berkman LF, Charpentier PA, Blazer DG: Behavioral and psychosocial pre-dictors of physical performance: MacArthur studies of successful aging. J Gerontol A Biol Sci Med Sci 1995; 50: 177–183.
21 Newsom JT, Nishishiba M, Morgan DL, Rook KS: The relative importance of three domains of positive and negative social ex-changes: a longitudinal model with compa-rable measures. Psychol Aging 2003; 18: 746–754.
22 Kiecolt-Glaser JK, Malarkey WB, Chee MA, Newton T, Cacioppo JT, Mao HY, Glaser R: Negative behavior during marital conflict is associated with immunological down-regu-lation. Psychosom Med 1993; 55: 395–409.
23 Cohen S, Tyrrell DAJ, Smith AP: Psychologi-cal stress and susceptibility to the common cold. N Engl J Med 1991; 325: 606–612.
24 Coyne JC, Downey G: Social factors and psy-chopathology: stress, social support, and coping processes. Annu Rev Psychol 1991; 42: 401–425.
25 Cohen S: Social relationships and health. Am Psychol 2004; 59: 676–684.
26 Sorkin D, Rook KS: Dealing with negative social exchanges in later life: coping respons-es, goals, and effectiveness. Psychol Aging 2006; 21: 715–725.
27 Lazarus RS, Folkman S: Stress, Appraisal, and Coping. New York, Springer, 1984.
28 Lakey B, Tardiff TA, Drew JB: Negative so-cial interactions: assessment and relations to social support, cognition, and psychological distress. J Soc Clin Psychol 1994; 13: 42–62.
29 Taylor SE, Klein LC, Lewis BP, Gruenewald TL, Gurung RA, Updegraff JA: Biobehavior-al responses to stress in females: tend-and-befriend, not fight-or-flight. Psychol Rev 2000; 107: 411–29.
30 Wethington E, Kessler RC: Perceived sup-port, received support, and adjustment to stressful life events. J Health Soc Behav 1986; 27: 78–89.
31 Thoits P: Stress, coping, and social support processes: where are we? What next? J Health Soc Behav 1995; 35: 53–79.
32 Walen HR, Lachman ME: Social support and strain from partner, family, and friends: costs and benefits for men and women in adulthood. J Soc Pers Relat 2000; 17: 5–30.
33 Carstensen LL: Social and emotional pat-terns in adulthood: support for socioemo-tional selectivity theory. Psychol Aging 1992; 7: 331–338.
34 Birditt K, Fingerman K: Do age differences in close and problematic family ties reflect the pool of available relatives? J Gerontol B Psychol Sci Soc Sci 2003; 58: 80–87.
35 Blanchard-Fields F: Everyday problem solv-ing and emotion: an adult developmental perspective. Curr Dir Psychol Sci 2007; 16: 26–31.
36 Krause N: Social support; in Binstock RH (ed): Handbook of Aging and the Social Sci-ences, ed 5. San Diego, Academic Press, 2001, pp 272–311.
37 Wenger GC: Social networks and gerontol-ogy. Rev Clin Gerontol 1996; 6: 285–293.
38 Kessler RC, Essex M: Marital status and de-pression: the importance of coping resourc-es. Soc Forces 1982; 61: 484–507.
39 Kiecolt-Glaser JK, Newton TL: Marriage and health: his and hers. Psychol Bull 2001; 127: 472–503.
Positive and Negative Social Exchanges Gerontology 13
40 van Tilburg T, van Groenou MB: Network and health changes among older Dutch adults. J Soc Issues 2002; 58: 697–713.
41 Unger JB, McAvay G, Bruce ML, Berkman L, Seeman T: Variation in the impact of social network characteristics on physical func-tioning in elderly persons: MacArthur stud-ies of successful aging. J Gerontol Series B 1999; 54:S245–S251.
42 Blazer D, Burchett B, Service C, George LK: The association of age and depression among the elderly: an epidemiologic exploration. J Gerontol A Biol Sci Med Sci 1991; 46:M210–M215.
43 Smith J, Borchelt M, Maier H, Jopp D: Health and well-being in the young old and oldest old. J Soc Issues 2002; 58: 715–732.
44 Australian Bureau of Statistics (ABS): Age by Sex, Census Tables, Census of Population and Housing 2006, Cat. No. 2068.0. Canber-ra, Australian Bureau of Statistics, 2007.
45 Pilkington PD, Windsor TD, Crisp DA: Vol-unteering and subjective well-being in mid-life and older adults: the role of supportive social networks. J Gerontol Series B 2012; 67: 249–260.
46 Lusas N, Windsor TD, Caldwell TM, Rod-gers B: Psychological distress in non-drink-ers: associations with previous heavy drink-ing and current social relationships. Alcohol Alcohol 2009; 45: 95–102.
47 Hays RD: R36 H.S.I. RAND 36 – 36 Health Status Inventory. Orlando, The Psychologi-cal Corporation/Harcourt Brace & Compa-ny, 1998.
48 Ware JE, Sherbourne CD: The MOS 36-Itcm Short-Form Health Survey (SF-36): I. Con-ceptual framework and item selection. Med Care 1992; 30: 473–483.
49 Aiken LS, West SG: Multiple Regression: Testing and Interpreting Interactions. New-bury Park, Sage, 1991.
50 Maxwell SE, Delaney HD: Designing Exper-iments and Analyzing Data: A Model Com-parison Perspective. Mahwah, Lawrence Erl-baum, 2004.
51 Swain SO: Men’s friendships with women: intimacy, sexual boundaries, and the infor-mant role; in Peter M (ed): Men’s Friend-ships: Research on Men and Masculinities. Newbury Park, Sage, 1992, vol 2, pp 153–171.
