Introduction

The concept of subjective aging experiences, that is, the way people perceive and evaluate their own aging process, has recently gained increased attention (Diehl et al. 2014). A rather novel construct in this context is the construct of awareness of age-related change (AARC). AARC “[…] refers to all those experiences that make a person aware that his or her behavior, level of performance, or ways of experiencing his or her life have changed as a consequence of having grown older […]” (Diehl and Wahl 2010). AARC captures perceived age-related gains (AARC-Gains) and losses (AARC-Losses) across different behavioral domains. The multidirectional and multidimensional character anchors the AARC construct in established life-span developmental psychology principles (Baltes et al. 2006) and distinguishes it from other subjective aging operationalizations, such as felt age (Diehl et al. 2014).

In prior research, AARC has been linked to depressive symptoms (Dutt et al. 2016). Depressive symptoms comprise sadness or absence of happiness and optimism (Radloff 1977). They naturally operate on the basis of a continuum from no-to-severe symptoms (Bowins 2015). In a similar vein, Siddaway et al. (2017) found that depressive symptoms constitute a single factor with a well-being/depression continuum.

More specifically, higher AARC-Losses were associated with an increase in depressive symptoms over 2.5 years (Dutt et al. 2016). Yet, it remains unclear whether this association unfolds over longer time frames and through which mechanisms AARC-Losses translate into depressive symptoms (i.e., mediators). Depressive symptoms reveal considerable frequency in middle and late adulthood (Luppa et al. 2012) and exert adverse effects on health (Ho et al. 2014) and mortality (Gallo et al. 2005). Given this key role of depressive symptoms in developmental psychology, we argue that additional research is needed to shed light on (pathways in) the AARC-Losses–depressive symptoms association. Drawing upon the work of Dutt et al. (2016), a first aim was to replicate the finding that AARC-Losses are associated with depressive symptoms in middle and old adulthood, relying on a 4.5-year interval. A second aim was to examine the role of mediators in this association.

Fundamental longitudinal associations between awareness of age-related losses and depressive symptoms

Theoretical considerations suggest bidirectional associations between AARC-Losses and depressive symptoms based on behavioral, cognitive, or motivational pathways. A detailed description of these mechanisms can be found in Dutt et al. (2016). For example, the perception of many age-related losses may be associated with less adaptive coping efforts which in turn should increase depressive symptoms (Wurm et al. 2013). Conversely, the negative developmental changes that are elicited by depressive symptoms (e.g., cognition; Bunce et al. 2014) may be misattributed to increasing age, such that AARC might be tainted by depressive symptoms. Both directions of effect can also be explained by mechanisms of selective attention and interpretation due to a desire for a consistent self-view (Swann 1983). Due to the deep incorporation of AARC into a person’s self-concept (Diehl et al. 2015), and in line with Dutt et al. (2016), we assumed that the effect of AARC-Losses on depressive symptoms is stronger than the converse. In a next step, we examined specific mediators in the AARC-Losses–depressive symptoms association.

Selected mediators in the association between awareness of age-related losses and depressive symptoms

In her stereotype embodiment theory, Levy (2009) proposes several pathways via which subjective aging experiences translate into developmental outcomes. Precisely, Levy suggests that subjective aging experiences exert an effect on well-being through psychological, behavioral, or physiological pathways. Given the specific nature of our data (i.e., questionnaire data), we decided to focus on psychological pathways in this work.

The AARC construct implies that perceived behavioral changes are attributed to increasing age (Diehl and Wahl 2010). Two considerations can be derived. First, AARC comprises temporal comparison processes, where past and ongoing experiences are integrated. It seems worthwhile to investigate how temporal comparison processes revolving around past experiences extend to the anticipation of future events. We therefore chose future time perspective (FTP) as a mediator in the AARC-Losses–depressive symptoms association. Second, AARC is deeply rooted in daily behavior, performance, and experiences. It seems necessary to examine how such age-related constraints in everyday life influence an individual’s expectations to efficiently deal with daily hassles. We therefore chose general self-efficacy (GSE) as a second mediator in the AARC-Losses–depressive symptoms association.

There is already evidence on a mediating role of FTP and GSE in the subjective aging–well-being association (Brothers et al. 2016; Stephan et al. 2011). These studies focused on life satisfaction and eudaimonic well-being; it remains unclear whether findings also apply to depressive symptoms. Moreover, previous studies were in part cross-sectional, thus prohibiting causal conclusions.

Future time perspective as a mediator in the association between awareness of age-related losses and depressive symptoms

FTP captures an individual’s perception of his or her remaining lifetime. With growing age, an individual perceives future time as being increasingly limited (Lang and Carstensen 2002).

We assumed a mediating role of FTP in the AARC-Losses–depressive symptoms association. First, AARC-Losses should impact FTP. Perceiving negative age-related changes might remind of being in the last phase of life and hence should narrow FTP (Brothers et al. 2016). Indeed, higher AARC-Losses are associated with a limited FTP (Brothers et al. 2016). This association also applies for other subjective aging measures. Older adults who psychologically dissociated themselves from their age group perceived their future as more open-ended (Weiss and Lang 2012).

Second, FTP should be related to depressive symptoms. A more limited FTP is associated with higher depressive symptoms (Hoppmann et al. 2017). This association might be explained by behavioral pathways, such as lower use of adaptive coping (Baltes et al. 2014), less positive social exchanges (Windsor et al. 2011), and less health behaviors (Stahl and Patrick 2012).

In the light of these findings, we assumed that high levels of perceived age-related losses narrow FTP, which in turn should increase the level of depressive symptoms.

General self-efficacy as a mediator in the association between awareness of age-related losses and depressive symptoms

GSE (Bandura 1977) captures the belief in one’s competence to tackle difficult or novel tasks and to cope with adversity in challenging situations (Luszczynska et al. 2005). GSE tends to decline from age 50 or 60 on (Schieman and Campbell 2001).

We assumed a mediating role of GSE in the AARC-Losses–depressive symptoms association. First, AARC-Losses should impact GSE. Perceiving loss experiences across different behavioral domains and attributing them to age, that is, to an internal factor, can be understood as absence of mastery experiences and thus should dampen self-efficacy (Bandura 1977). The association between subjective aging experiences and GSE has also been shown empirically. People who feel younger than their age are more likely to experience high GSE (Teuscher 2009).

Second, GSE is related to depressive symptoms (Steunenberg et al. 2006). This association can be explained by higher well-being-enhancing behaviors (e.g., health behaviors, adaptive coping) in individuals with high self-efficacy beliefs (Bandura 1977; Jang et al. 2002).

Based on these considerations, we assumed that high levels of perceived age-related losses dampen GSE, which in turn should increase the level of depressive symptoms.

Research aims and hypotheses

This study examined the association between AARC-Losses and change in depressive symptoms in middle-aged and older adults using three measurement points covering 4.5 years. First, extending Dutt et al.’s (2016) finding, we assumed that the effect of AARC-Losses on change in depressive symptoms over 4.5 years is stronger than the reverse direction (Hypothesis 1). Second, we expected that the detrimental effect of AARC-Losses on change in depressive symptoms over 4.5 years is mediated by FTP and GSE (Hypothesis 2). Theoretical considerations suggest changes in the normativity of age-related experiences across the life span, with older adults having a higher likelihood of experiencing losses (Baltes et al. 1980). Therefore, we also exploratorily tested for age effects by contrasting the hypothesized pattern of associations between middle-aged (< 65 years) and old-aged (≥ 65 years) adults.

Method

Study design and participants

Data were collected in 2012 (T1), 2015 (T2), and 2017 (T3). Ethical approval for T3 was obtained from the research ethics committee of the Faculty of Behavioral and Cultural Studies at Heidelberg University. The observational interval covered on average 2.56 years (SD = 13.18 days) between T1 and T2 and 4.67 years (SD = 13.63 days) between T1 and T3.

In total, 423 persons participated at T1, 356 persons participated at T2 (84%), and 299 persons participated at T3 (71%). Main reasons for dropout from T1 to T2 were no interest (34%), health problems (9%), or death (6%). Main reasons for dropout from T1 to T3 were no interest (55%), health problems (7%), or death (10%). Regarding the T2 measurement, dropouts were older than participants, d = 0.49, and reported a worse physical health, d = − 0.25. Regarding the T3 measurement, dropouts were more common among men than among women. Dropouts were older than participants, d = 0.41. Participants and dropouts for T2 and T3, respectively, did not differ on the remaining study variables at T1, all ps > .05. Thus, analyses were based on the 423 T1 participants. Missing data were replaced using the expectation–maximization (EM) algorithm, such that the final sample size amounted to N = 423. The EM algorithm provides robust estimates in the reconstruction of the missing sample data (Gold and Bentler 2000).

At T1, participants were between 40 and 98 years old (M = 62.94, SD = 11.84 years). Two-thirds were women (64%). 61% were partnered or married. Education was high, with an average of 11.53 years (SD = 1.96 years) of schooling, and with more than half of the participants (55%) having a general qualification for university entrance. 40% were retired. 84% reported an average or above-average subjective health. Descriptive information of the study variables is summarized in Table 1.

Table 1 Descriptive information and bivariate associations between the study constructs, assessed at T1 (2012), T2 (2015), and T3 (2017)
Full size table

Measures

Predictor: awareness of age-related losses

AARC-Losses were measured with the 25-item loss-related subscale of the AARC questionnaire (Brothers et al. 2018). Each item is preceded by the item stem “With my increasing age, I realize that…” and comprises the description of a negative experience in one out of five behavioral domains, that is, health and physical functioning, cognitive functioning, interpersonal relations, social-cognitive/social-emotional functioning, and lifestyle and engagement. The 25 items evenly represent the five behavioral domains. An example for a negative aging experience is “…that it is more difficult for me to learn new things.” Participants indicated their agreement on a 5-point Likert scale, from 1 (not at all) to 5 (very much). The internal consistency for the scale was good (Cronbach’s α = .92 at T1 and .93 at T3). For latent modeling, we parceled the indicator variables (Little et al. 2002). The five domain mean scores served as indicators for the AARC-Losses factor. We calculated a composite reliability index as an indicator of the internal consistency of the scale (Raykov 1997), which is estimated based on the parcel factor loadings. This composite reliability amounted to .86 at T1 and .89 at T3.

Outcome: depressive symptoms

The Center for Epidemiologic Studies Depression Scale (CES-D; Radloff 1977) measures an individual’s current level of depressive symptoms. The CES-D does not assess clinical symptoms of depression, but constitutes a well-being/depression continuum (Siddaway et al. 2017). The 10-item version used here shows good psychometric properties in old adults (Irwin et al. 1999). The items were evaluated on the basis of the symptoms during the past week and judged on a 4-point Likert scale, from 1 (rarely or none of the time) to 4 (most or all of the time). The internal consistency was good (Cronbach’s α = .85 at T1 and .86 at T3). For latent modeling, items were parceled into three parcels comprising 3–4 items, with an equal distribution of factor loadings across parcels. The composite reliability amounted to .86 at T1 and T3.

Mediator: future time perspective

FTP was operationalized by a 10-item scale developed by Carstensen and Lang (1996). An example for an item formulation is “My future seems infinite to me.” The items were evaluated on a 7-point Likert scale, from 1 (very untrue) to 7 (very true). The scale yielded a good internal consistency (Cronbach’s α = .89 at T1 and T2 and .86 at T3). For latent modeling, items were parceled into three parcels comprising 3–4 items with an equal distribution of factor loadings. The composite reliability amounted to .91 at T1, .90 at T2, and .88 at T3.

Mediator: general self-efficacy

GSE was assessed using the General Self-Efficacy Scale by Schwarzer and Jerusalem (1995). Ten items assess a person’s confidence to overcome difficult situations due to his or her own competence, for example “I am confident that I could deal efficiently with unexpected events.” Each statement was judged on a 4-point Likert scale, from 1 (not at all true) to 4 (exactly true). The internal consistency was good (Cronbach’s α = .91 at T2 and T3; GSE was not assessed at T1). For latent modeling, items were parceled into three parcels comprising 3–4 items with an equal distribution of item factor loadings. The composite reliability amounted to .91 at T2 and .92 at T3.

Control variables

T1 calendar age, sex, education, physical health, and neuroticism were used as manifest control variables, as has been done in Dutt et al. (2016). Sex was coded as 0 (male) and 1 (female). Education was operationalized as years of schooling. Physical health was operationalized as the SF-36 physical health component summary score, which comprises eight subscales (Bullinger et al. 1995). Cronbach’s α ranged between .76 and .93 for the eight SF-36 subscales. Neuroticism was assessed by the NEO five-factor inventory (Borkenau and Ostendorf 2008). The items were rated on a 5-point Likert scale, ranging from 1 (strongly disagree) to 5 (strongly agree). The neuroticism subscale showed a good internal consistency (Cronbach’s α = .84).

Data analyses

Hypothesis 1

To test our first hypothesis, that is, AARC-Losses exert an effect on change in depressive symptoms, we specified a latent cross-lagged panel model realized with AMOS 24. The structural model included four latent variables, that is, AARC-Losses and depressive symptoms, both assessed at T1 and T3 (Fig. 1). The control variables were correlated with T1 AARC-Losses and T1 depressive symptoms. We first specified a model with the autoregressive paths only. This model was compared in terms of χ2 to a model including the autoregressive paths and all cross-lagged paths (full model). Cross-lagged path coefficients that turned out to be nonsignificant in the full model were restricted to 0, and change in χ2 was evaluated.

Fig. 1
figure 1

Structural equation model examining associations between Time 1 (T1, 2012) and Time 3 (T3, 2015) awareness of age-related losses (AARC-Losses) and depressive symptoms (DEP). Notes. e manifest variable residual; I manifest indicators modeled as item parcels; r latent variable residual. The analyses controlled for T1 age, sex, education, physical health, and neuroticism. Residual paths are constrained to 1. All parameters reported have been adjusted by the expectation–maximization (EM) estimate. N = 423. ***p < .001

Full size image

Hypothesis 2

To test our hypothesis on a mediating role of FTP and GSE in the AARC-Losses–depressive symptoms association, we specified a latent longitudinal mediation model. As mediation consists of causal processes that unfold over time, longitudinal mediation models account for the temporal structure that is required to test mediation, thus constituting a clear advantage over cross-sectional mediation models (Maxwell and Cole 2007). The structural model resembled the model tested within the context of Hypothesis 1, but was extended by the two mediators. In order to account for time-synchronous associations between the mediators and the predictor and the outcome, respectively, assessments of the mediators were included for the three measurement points (except for T1 GSE, as GSE had only been measured at T2 and T3). The model thus also takes into consideration the stability of the study variables across the observational interval (Cole and Maxwell 2003). Paths were specified from T1 AARC-Losses to T3 depressive symptoms, from T1 AARC-Losses to T2 FTP and GSE, and from T2 FTP and GSE to T3 depressive symptoms (mediation model in its narrower sense). We specified a covariance between the residuals of T2 FTP and GSE (Luszczynska et al. 2005) (Fig. 2). Again, the control variables were correlated with all T1 variables.

Fig. 2
figure 2

Structural equation model examining the mediation of Time 1 (T1, 2012) awareness of age-related losses (AARC-Losses) on Time 3 (T3, 2017) depressive symptoms (DEP) via Time 2 (T2, 2015) future time perspective (FTP) and T2 general self-efficacy (GSE). Notes. e manifest variable residual; I manifest indicators modeled as item parcels; r latent variable residual. The analyses controlled for T1 age, sex, education, physical health, and neuroticism. Residual paths are constrained to 1. Autoregressive path coefficients are printed in italics. Path coefficients of the mediation model in its narrowest sense are printed in bold. The total effect is reported in brackets. All parameters reported have been adjusted by the expectation–maximization (EM) estimate. N = 423. †p < .10, *p < .05, **p < .01, ***p < .001

Full size image

To test the mediating role of FTP and GSE for the AARC-Losses–depressive symptoms association, the significance of the indirect effect was assessed using 95% bias-corrected bootstrapped confidence intervals (BCCI) based on 5000 bootstrap samples (Preacher and Hayes 2008). We report κ2 as an effect size measure (Preacher and Kelley 2011). κ2 can be interpreted as the proportion of the maximum possible indirect effect that could have occurred (range 0–1). We also report the partially standardized indirect effect abps, that is, the ratio of the indirect effect to the standard deviation of the outcome (Preacher and Kelley 2011). abps indicates by how many standard deviations the outcome changes for every one-unit increase in the predictor indirectly via the mediator.

Results

Bivariate correlations between the study constructs are displayed in Table 1. T1 AARC-Losses were positively linked to depressive symptoms within and across measurements. AARC-Losses and depressive symptoms were negatively linked to FTP and GSE within and across measurements.

Measurement invariance

Measurement invariance across time

As AARC-Losses, depressive symptoms, FTP, and GSE were assessed at different occasions, we tested for temporal measurement invariance (MI) by specifying progressively stricter measurement models (Horn and McArdle 1992). For all constructs, modification indices suggested a correlation between corresponding errors across measurements (e.g., T1 depressive symptoms parcel 1 error with T3 depressive symptoms parcel 1 error) that we admitted. Following the recommendations of Cheung and Rensvold (2002), a decrease by > .01 in the comparative fit index (CFI) after imposing restrictions to the model was interpreted as a violation of MI. As we did not estimate parcel intercepts within the models, weak MI (i.e., equal factor loadings across measurements) is sufficient to proceed with the main hypothesis testing. Using the criterion by Cheung and Rensvold (2002), weak MI could be established for all study constructs.

Measurement invariance across age groups

In order to investigate age effects between middle-aged (< 65 years; N = 255) and older (≥ 65 years; N = 168) adults, we tested for MI across age groups. We chose age 65 as cutoff for the age groups, as this age is tied to legislative regulations in Germany, such as entry into retirement, thus separating two life phases. Participants in the old-age group were older, d = 2.69, reported worse physical health, d = −0.38, more AARC-Losses, T1: d = 0.47; T3: d = 0.51, and a more limited FTP, T1: d = −0.88, T2: d = −0.69, T3: d = −0.66, as compared to participants in the middle-age group. Men were more strongly represented in the old-age group as compared to the middle-age group. Using ∆CFI as a criterion, weak MI could be established across age groups for all study variables.

Hypothesis 1: findings of a cross-lagged panel model linking AARC-Losses with depressive symptoms over 4.5 years

The model containing all cross-lagged paths fit the data better than the model containing the autoregressive paths only, ∆χ2(2) = 18.04, p < .001. A closer inspection of the cross-lagged path coefficients revealed that the path from T1 depressive symptoms to T3 AARC-Losses was not significant, p = .188. Restraining this path to 0 did not significantly weaken the model fit, ∆χ2(1) = 1.67, p = .196. The reduced model yielded acceptable fit, χ2(162) = 584.24, CFI = .927, root mean square error of approximation (RMSEA) = .079, 90% confidence interval (CI) [.072, .086]. The effect of T1 AARC-Losses on T3 depressive symptoms was significant, b = 0.17, standard error (SE) = 0.04, p < .001. The effect size, calculated as z-value (b/SE) relative to the root of the sample size (Rosenthal 1994), amounted to r = .21, indicating a small effect. Hence, higher T1 AARC-Losses were associated with an increase in depressive symptoms over 4.5 years, whereas the reverse direction of effects was not significant (Fig. 1). The result thus extends the findings from Dutt et al. (2016) to a 4.5-year interval.

We tested for differences between middle-aged and older adults. Based on the assumption of weak MI across age groups, we imposed equality constraints on the cross-lagged and autoregressive paths across age groups within the full model. Compared with the unconstrained model, there was no significant worsening of model fit, ∆χ2(4) = 6.27, p = 0.180. Hence, it can be assumed that the effects were equivalent across age groups.

Hypothesis 2: findings of a longitudinal mediation model targeting future time perspective and general self-efficacy

The longitudinal mediation model yielded a satisfactory fit, χ2(541) = 1542.27, CFI = .918, RMSEA = .066, 90% CI [.062, .070]. The model explained 46.2% of variance in depressive symptoms (Fig. 2). The direct effect of T1 AARC-Losses on T2 FTP was significant, b = −0.24, SE = 0.06, p < .001, as was the direct effect of T1 AARC-Losses on T2 GSE, b = −0.23, SE = 0.03, p < .001. T2 FTP and GSE, in turn, exerted a (marginally) significant direct effect on T3 depressive symptoms, b = −0.05, SE = 0.03, p = .056 for FTP; b = −0.13, SE = 0.05, p = .011 for GSE. Hence, the perception of one’s aging as associated with many losses leads to a reduction in perceived future time and general self-efficacy 2.5 years later, which in turn intensified depressive symptoms a further 2 years later, controlling for age, sex, education, physical health, and neuroticism, and taking into account the autoregressive effects of the study variables over time as well as their time-synchronous correlations.

The total effect of T1 AARC-Losses on T3 depressive symptoms was significant, b = 0.15, SE = 0.05, p = .002. When the mediators were included, the direct effect of T1 AARC-Losses on T3 depressive symptoms was attenuated, b = 0.11, SE = 0.05, p = .024. The total indirect effect of T1 AARC-Losses on T3 depressive symptoms through the two mediators amounted to significance, b = 0.04, SE = 0.02, 95% BCCI [0.02, 0.08], p < .001. This suggests that the effect of AARC-Losses on depressive symptoms was mediated via FTP and GSE. The complete indirect effect can be divided into components attributable to the mediational effect of FTP and GSE, respectively. The two specific indirect effects amounted to significance each, with b = 0.01, SE = 0.01, 95% BCCI [0.00, 0.03], p = .018, abps = 0.02, κ2 = .02, for FTP, and b = 0.04, SE = 0.01, 95% BCCI [0.01, 0.07], p = .001, abps = 0.07, κ2 = .06, for GSE. The difference between the two specific indirect effects was not significant, p = .589.

We next examined the specific indirect effects separately for each mediator within two single mediator models, with the other mediator removed from the model. The indirect effect through FTP amounted to significance, b = 0.02, SE = 0.01, 95% BCCI [0.00, 0.04], β = .11, p = .003, abps = 0.03, κ2 = .02. Regarding GSE as a mediator, the indirect effect was significant, too, b = 0.04, SE = 0.01, 95% BCCI [0.01, 0.07], β = .05, p = .002, abps = 0.07, κ2 = .06. Hence, FTP and GSE, considered both in combination as well as alone, can be viewed as significant mediators in the AARC-Losses–depressive symptoms association.

We tested for differences between middle-aged and older adults. Based on the assumption of weak MI across age groups, we imposed equality constraints on the structural paths across age groups. Compared with the unconstrained model, there was no significant worsening of model fit, ∆χ2(10) = 17.63, p = .062. Hence, it can be assumed that the effects were equivalent across age groups.

Discussion

We were able to replicate the finding on an effect of AARC-Losses on change in depressive symptoms, controlling for T1 depressive symptoms, in middle- and old-aged adults that has been obtained by Dutt et al. (2016). An extension of this study was the inclusion of a 4.5-year interval. We next examined FTP and GSE as mediators in the AARC-Losses–depressive symptoms association. Many perceived losses dampened FTP and GSE 2.5 years later, which in turn increased depressive symptoms a further 2 years later. Effects were robust after controlling for age, sex, education, physical health, and neuroticism, and taking into account the stabilities and time-synchronous correlations of the study variables. Effects were comparable in middle- and old-aged adults.

Awareness of age-related losses is associated with change in depressive symptoms

In line with Dutt et al. (2016), many perceived age-related losses were associated with an increase in depressive symptoms over 4.5 years, whereas the reverse direction of effect did not reach significance. The independence of AARC-Losses from depressive symptoms points to the robustness of AARC and suggests that AARC might be deeply incorporated into a person’s self-concept (Diehl et al. 2015), thus providing little target for depressive symptoms.

Levy (2009) assumes within her stereotype embodiment theory that the effect of subjective aging experiences on well-being may be mediated through psychological, behavioral, and physiological mechanisms. We focused on the psychological pathway in this work and examined FTP and GSE as mediators in the AARC-Losses–depressive symptoms association. By drawing on three measurement points, we were able to address the chronological sequence of AARC-Losses, FTP, GSE, and depressive symptoms.

Future time perspective and general self-efficacy mediate the association between awareness of age-related losses and depressive symptoms

The mediating role of FTP and GSE in the AARC-Losses–depressive symptoms association can be illustrated exemplarily with regard to age-related loss experiences in the physical domain (i.e., increased bodily pain). Such an experience is particularly detrimental as the perceived losses are attributed to age, that is, to an internal and irreversible condition (Diehl and Wahl 2010). An individual who is aware of negative physical changes might interpret them as a sign of the aging process constantly moving ahead. In the light of this immediate confrontation with one’s life finitude, his or her FTP is dampened (Brothers et al. 2016). Moreover, a person who encounters many physical declines accompanied by constraints in everyday life might start questioning his or her competence to efficiently deal with daily hassles, attenuating GSE (Bandura 1977). The reduced FTP and GSE should turn the individual into a state of avoidance, refraining from health behaviors (e.g., Stahl and Patrick 2012), social contacts (e.g., Windsor et al. 2011), or adaptive coping (e.g., Baltes et al. 2014). At the end of the day, the individual will undergo an increase in depressive symptoms.

We exclusively focused on the loss component of AARC. This decision was driven by the observation that AARC-Losses were associated with depressive symptoms, whereas AARC-Gains were unrelated to depressive symptoms in prior research (Dutt et al. 2016). This is in line with Baumeister et al.’s (2001) general observation that negative events exert a stronger impact on developmental outcomes as compared to positive events (see also Meisner 2012). We exploratorily repeated our analyses with AARC-Gains. The total effect of AARC-Gains on depressive symptoms was not significant over the 4.5-year interval, b = 0.03, SE = 0.04, p = .505. Moreover, the indirect effect through FTP and GSE did not reach significance, b = −0.01, SE = 0.02, 95% BCCI [−0.04, 0.02], β = –.02, p = .684. Hence, the mediating role of FTP and GSE that we found in the AARC-Losses–depressive symptoms association cannot be generalized to AARC-Gains.

We did not find any age effects regarding the patterns of association. It has to be taken into account that our sample included few very old adults (only 11.1% were 80 years and older at T1). The inclusion of more adults in advanced old age might have increased the probability of finding age effects.

At the practical level, our results can contribute to the prevention and treatment of depressive disorders in the second half of life. Although the treatment of depressive symptoms requires interdisciplinary therapeutic approaches, a better understanding of the mechanisms underlying the harmful association between perceived age-related losses and depressive symptoms could provide a starting point for the development of interventions to overcome this detrimental relationship. When one is sensitive to, for example, the potentially hazardous effect of high AARC-Losses on GSE, a therapeutic goal could be to purposefully provide the aging individual with experiences of mastery to maintain high levels of GSE despite many perceived losses. Similarly, the detrimental association between AARC-Losses and FTP can be attenuated by redirecting older adults’ view of the future from a focus on limitations to a focus on opportunities (Cate and John 2007) (e.g., via increasing the awareness of gain experiences or via a reformulation of long-term goals to short-term goals that can be attained in the near future). Future studies are needed to tailor such intervention programs.

Limitations and future research

The sample was positively biased with regard to education or health. As such, caution is advisable when generalizing the results to other populations.

Although we controlled for T1 depressive symptoms and FTP, thus depicting change over 4.5 years, GSE was only assessed at T2 and T3. Future investigations should examine how AARC-Losses predict change in GSE.

All effects were of small effect size. However, given the high practical significance of depressive symptoms, also effects of a small size merit the attention to be investigated (Roberts et al. 2007).

By choosing FTP and GSE as mediators, we took into consideration two psychological mechanisms as postulated by stereotype embodiment theory (Levy 2009). Notwithstanding, other constructs, for example, coping, health (behaviors), physiological or physical activity, are also conceivable as mediators and should be investigated in future research (see also Stephan et al. 2011).

Theoretical considerations suggest that individual dispositions may amplify or buffer the effect of subjective aging experiences on depressive symptoms (Dutt et al. 2018; Wurm and Benyamini 2014), such that FTP and GSE could also function as a moderator in the AARC-Losses–depressive symptoms association. In additional analyses, however, the interaction between T1 AARC-Losses and FTP in the prediction of change in depressive symptoms did not reach significance. As GSE was not assessed at T1, we were not able to also test such a moderator hypothesis for GSE. This should be on the agenda for future research.

In conclusion, we found that the harmful effect of many perceived age-related losses on depressive symptoms over 4.5 years was mediated by FTP and GSE, controlling for demographic variables, physical health, and neuroticism. Our results suggest that the translation of subjective aging experiences into developmental outcomes is a complex and multistage phenomenon.