Self-esteem and social interactions in daily life: An experience sampling study

Abstract

Theory and empirical evidence suggest that self-esteem and the quality of social relationships are mutually related, however, little is known about the dynamic short-term association in daily life. Experience sampling data of 232 Dutch Master’s students (M _age = 24.6, SD _age = 2.7) was collected across four waves; each consisting of 14 days of four daily assessments (23,107 data points). We applied Dynamic Structural Equation Modeling to study the within-day associations between momentary self-esteem (assessed in the morning and evening) and social interactions (anticipated and experienced). First, results showed that quality of social interactions during the day was associated with evening, but not morning self-esteem. Second, positive anticipation of social interactions was associated with morning, but not with evening self-esteem. Third, trait self-esteem did not moderate the within-day associations. Together, results suggest that when having higher self-esteem, people anticipate more positive social interactions at the same time, but they do not experience more positive social interactions later on. Having but not anticipating more positive social interactions seems to contribute to higher self-esteem later on. Differentiating between anticipated and experienced interactions and using multiple daily assessments helps obtain a more fine-grained understanding of the dynamic processes between self-esteem and social interactions.

Plain language summary

Social interactions, such as having a laugh with a friend or a quibble with a partner: we have many every day. Interestingly, research shows that the quality of these social interactions—and our social relationships overall—are connected to how much we value ourselves, known as self-esteem. In our study, we investigated how social interactions and self-esteem are linked throughout the day. We collected data from 232 Dutch students, who answered short surveys on their smartphones four times a day over 2 weeks. We found three key results: First, we found that the quality of social interactions during the day was related to higher self-esteem in the evening, but not in the morning. Second, when expecting positive social interactions during the day, people tended to have high self-esteem in the morning. However, this wasn’t true for evening self-esteem. Third, we also examined if a person’s overall self-esteem (trait self-esteem) played a role in how these relationships worked. It didn’t—the links between daily social interactions and self-esteem were irrespective of trait self-esteem. In short, when having higher self-esteem, people seem to expect better social interactions at the same time, but it does not seem to boost better social interactions later in the day. However, having, but not expecting better social interactions seems to boost self-esteem later on. In summary, this study helps us better understand how social interactions and self-esteem influence each other in real time. We hope future research will explore these relationships over shorter periods (within hours) and with more diverse groups beyond students.

Introduction

Social relationships are essential for our everyday happiness and well-being (Leary & Baumeister, 2000). The quality of social relationships has been linked to self-esteem, which is a person’s evaluation of their worth as a person (Trzesniewski et al., 2013). It is well grounded in theory that self-esteem and the quality of social relationships mutually influence each other. According to sociometer theory, self-esteem serves as a measure to indicate one’s relational value (Leary & Baumeister, 2000). This suggests that the quality of social relationships can influence one’s self-esteem. Vice versa, approaches such as the self-broadcasting perspective suggest that people’s self-esteem affect their social relationships (Srivastava & Beer, 2005; Zeigler-Hill et al., 2013). According to this perspective, people express their self-esteem in their social interactions, which may affect the quality of their social relationships.

The literature shows that self-esteem development is embedded within social environments and provides supporting evidence for prospective relationships between self-esteem and social relationship quality (for a large-scale data analysis, see de Moor et al. (2021); for a meta-analysis, see Harris and Orth (2020)). In daily life, self-esteem and daily manifestations of social relationships also seem to be intertwined. Multiple studies found concurrent associations between state or momentary self-esteem (hereafter momentary self-esteem) and interpersonal measures within the same day (Bleckmann et al., 2022; Denissen et al., 2008; Heppner et al., 2008; Mahadevan et al., 2020; Nezlek, 2005; Pietromonaco & Barrett, 2009; Schmidt et al., 2020; Subrahmanyam et al., 2020; Wagner et al., 2023). In line with these findings, daily diary studies looking into the prospective associations found support for interpersonal measures predicting daily self-esteem on the next day (Denissen et al., 2008; Murray, et al., 2003b; Subrahmanyam et al., 2020; Wagner et al., 2023), and less conclusive support for interpersonal measures predicting momentary self-esteem between and within days (Denissen et al., 2008; Murray, et al., 2003b; Schmidt et al., 2020).

The present study examines the association between momentary self-esteem (self-esteem in the moment) and the quality of social interactions in daily life, as done in previous research (Bleckmann et al., 2022; Denissen et al., 2008; Hutteman et al., 2015; Schmidt et al., 2020; Subrahmanyam et al., 2020; Wagner et al., 2023). Our study extends this research by assessing the prospective associations (instead of concurrent associations (Bleckmann et al., 2022; Wagner et al., 2023)) and across shorter time scales (within days instead of across days or months (Denissen et al., 2008; Hutteman et al., 2015; Subrahmanyam et al., 2020)). These within-day associations have been studied before by Schmidt et al. (2020) in a children sample, however, it remains unclear if the within-day processes found in their study apply to young adults as well. We replicated this research using four waves of experience sampling data that were analyzed within the Dynamic Structural Equation Modeling (DSEM) framework, which allowed for a further inspection of individual differences in the studied prospective effects. A main advantage of DSEM is the possibility to estimate person-specific effect sizes for the within-day associations. Linking these person-specific effects with trait measures can aid our understanding of individual differences in personality traits and their link to within-day processes of momentary self-esteem and quality of social interactions.

Next, we aimed to extend existing research by not only examining the link between momentary self-esteem (measured as morning/evening self-esteem) and quality of social interactions, but also the anticipation (positive or negative) of social interactions. In line with research showing that self-esteem can already change in the anticipatory phase before life events (see Reitz (2022)), it is imaginable that interpersonal perceptions in the anticipatory phase before daily social events can also initiate momentary self-esteem changes. Studies suggest that even a single social experience can affect momentary self-esteem (Bleckmann et al., 2022; Thomaes et al., 2010). This change in momentary self-esteem may not only be due to the actual social experience, but also to anticipatory processes that take place beforehand. Not only experiencing but also anticipating social interactions may initiate daily self-esteem processes.

Direct evidence is missing, but there is some indirect evidence. In the emotions literature, there is evidence suggesting that the mere anticipation of a stressor can already initiate physiological (Waugh et al., 2010) and affective responses (Neubauer et al., 2018), which underlines the importance of studying anticipation effects in addition to actual experiences. In the self-esteem literature, there is evidence suggesting that not only experiencing but also anticipating life events is linked to changes in self-esteem (Bleidorn & Schwaba, 2018; Reitz et al., 2020, 2022; van Scheppingen et al., 2018). Concerning self-esteem in daily life, daily feelings of concern about romantic partner’s acceptance and love (which could be considered a daily negative anticipation of acceptance by a romantic partner) have been shown to be linked to daily self-esteem sensitivity (Murray et al., 2003a). In line with this research, it is conceivable that a positive or negative anticipation of social interactions can either protect or threaten momentary self-esteem during the day.

Finally, we aimed to extend existing research by exploring if trait measures of self-esteem can explain heterogeneity in the within-day associations between momentary self-esteem and social interaction quality/anticipation. More insights into this top-down effect of trait self-esteem predicting heterogeneity in momentary self-esteem reactivity can ultimately aid our understanding of causal processes underlying self-esteem change, both on the short- and long-term (Reitz, 2022).

Momentary self-esteem and quality of social interactions (H1a, H1b)

In the present study, we aimed to understand how the prospective association between momentary self-esteem and social interaction quality unfolds within days in daily life (see Figure 1). Multiple studies have shown that momentary self-esteem is reactive to social interactions in concurrent associations (Bleckmann et al., 2022; Wagner et al., 2023), however, less is known about their cross-lagged effects as assessed in prospective associations (Wagner et al., 2023). Hence, we examined whether morning self-esteem predicted the quality of social interactions during the day – and whether the quality of social interactions during the day predicted evening self-esteem.

Figure 1.

Conceptual Model of Hypotheses 1a–1b and 3a.

Morning self-esteem predicting social interaction quality

In line with the self-broadcasting perspective, it is conceivable that morning self-esteem is positively associated with how individuals express themselves in social interactions during the day, which in turn positively relates to the quality of social interactions (Srivastava & Beer, 2005; Zeigler-Hill et al., 2013). From other theoretical accounts, however, can be derived that the effect of morning self-esteem on the quality of social interactions during the day might vary largely across individuals. For some individuals, low morning self-esteem might be a motivation to engage in protective behaviors in their relationships that improve the quality of their social interactions during the day (Murray et al., 2003b). For others, low morning self-esteem may trigger less constructive relationship patterns (Murray et al., 2003b), such as attributing their low morning self-esteem to low relational value. This may result in either an underestimation of the quality of social interactions (Downey & Feldman, 1996; Murray et al., 2003a; Murray et al., 2000) or an overreaction during social interactions and consequently lower quality of social interactions (Downey & Feldman, 1996; Murray et al., 2003b). Altogether, these theories highlight the different pathways through which morning self-esteem and quality of social interactions during the day may be associated within days.

Previous research showed inconsistent results regarding the effect of self-esteem on the quality of social relationships and interactions. Multiple longitudinal studies found supportive evidence for self-esteem predicting social relationship outcomes across months and years (de Moor et al., 2021; Harris & Orth, 2020; Hutteman et al., 2015; Reitz et al., 2016; Wagner et al., 2018). Zooming in on the daily life, daily self-esteem did not predict the quality of social interactions on subsequent days (Denissen et al., 2008; Subrahmanyam et al., 2020), except when a male interacted with their closest friend (Denissen et al., 2008). Using different measures, Murray et al. (2003b) found that state self-esteem predicted how much an individual felt rejected or accepted by their romantic partner during the subsequent day, indicating that momentary self-esteem predicts interpersonal perceptions across days. Moreover, there is preliminary evidence that self-esteem processes might be linked to quality of social interactions within days, as Schmidt et al. (2020) found that changes from morning to evening self-esteem were associated with social inclusion in a children sample. Given the different theoretical perspectives and mixed empirical support, we expected morning self-esteem to be associated with the quality of social interactions during the day but did not have a priori predictions about the valence of the effect.

Quality of social interactions predicting evening self-esteem

The reverse effect of quality of social interactions during the day on evening self-esteem is more well-founded. According to sociometer theory, self-esteem serves as an indicator for one’s relational value (Leary & Baumeister, 2000). This sociometer effect has been found for different operationalizations of relational value (i.e., social inclusion, acceptance/rejection, quality of social relationships and social interactions (Harris & Orth, 2020; Wagner et al., 2023)). In the present study, we focus on the aggregated quality score of all salient social interactions during the day. Importantly, the premise that social relationships shape self-esteem over time is rooted in many different theories (Ainsworth, 1973; Bowlby, 1973; Cooley, 1902; Harris & Orth, 2020; Harter, 1999; Mead, 1934; Shrauger & Schoeneman, 1979).

Longitudinal studies consistently showed that social inclusion and relationships shape self-esteem development over time (de Moor et al., 2021; Harris & Orth, 2020; Hutteman et al., 2015; Reitz et al., 2016; Wagner et al., 2018). Across months, Hutteman et al. (2015) found that state social inclusion predicted changes in state self-esteem. Between days, state interpersonal measures (i.e., quality of social interactions, perceived acceptance, perceived rejection) predicted daily self-esteem on the next day (Denissen et al., 2008; Murray et al., 2003b; Subrahmanyam et al., 2020). Within days, social inclusion during the day was positively related to evening self-esteem (Schmidt et al., 2020). In a laboratory setting, Thomaes et al. (2010) found that experimentally induced peer approval increased state self-esteem, while peer disapproval decreased state self-esteem. Based on these findings, we expected that the quality of social interactions during the day was positively associated with evening self-esteem (within days).

Momentary self-esteem and anticipated social interactions (H2a, H2b)

Momentary self-esteem may not only be linked with the actual experience of social interactions, but also with the positive or negative anticipation of social interactions (i.e., whether someone is looking forward to social interactions during the day or not). In line with literature showing that self-esteem changes can already occur in the anticipatory phase before life events (Reitz et al., 2022), it is imaginable that a positive or negative outlook in anticipation of daily social events is also linked to momentary self-esteem changes. We investigated the possible associations between momentary self-esteem and anticipated social interactions in daily life in two ways (see Figure 2). First, we tested the association between morning self-esteem and anticipation of social interactions assessed in the morning, to examine whether these constructs were associated before the actual social interactions took place. Second, we tested the association between evening self-esteem and anticipation of social interactions to examine whether evening self-esteem is reactive to the anticipation of social interactions assessed in the morning.

Figure 2.

Conceptual Model of Hypotheses 2a–2b and 3b.

Morning self-esteem and anticipated social interactions during the day

In the context of life events, it has been suggested that self-esteem can change in the anticipatory phase before the life event occurs (Bleidorn & Schwaba, 2018; Reitz et al., 2022; van Scheppingen et al., 2018). It remains unclear whether this effect also holds in the context of daily events. Individuals may increase in morning self-esteem when they anticipate positive social interactions during the day and decrease in morning self-esteem when they expect negative social interactions during the day. If so, this would imply that morning self-esteem is reactive to anticipation of social interactions during the day, even before actual social interactions take place. It is also conceivable that morning self-esteem shapes the anticipation of social interactions the other way around. Morning self-esteem can guide an individual in approaching social interactions during the day. For instance, low morning self-esteem might make individuals more aware of potential negative feedback/rejection in social interactions, while high morning self-esteem might make individuals more aware of positive feedback within social interactions (Leary & Baumeister, 2000; Swann & Buhrmester, 2012).

Previous research provides preliminary support for a within-day association between momentary self-esteem and interpersonal perceptions in daily life. A daily diary study found that state self-esteem on one day predicted the feelings of concern about the romantic partners’ acceptance on the subsequent day (Murray, Griffin, et al., 2003b). These results point towards the possibility that morning self-esteem might be linked to the anticipation of social interactions. We expected a positive association between morning self-esteem and the anticipation of social interactions.

Anticipated social interactions predicting evening self-esteem

The reverse effect of anticipation of social interactions predicting evening self-esteem is also conceivable. Individuals who expect to be rejected or accepted in social interactions may experience social interactions in line with these beliefs, which, in turn, can affect evening self-esteem. A more positive anticipation of social interactions may then be associated with higher evening self-esteem, while a negative anticipation may be associated with lower evening self-esteem. In an experimental study, Leary et al. (1998) showed that simply imagining social interactions in which rejection is involved can already lead to decreases in state self-esteem. Moreover, interpersonal perceptions such as feelings of concern about the acceptance from a romantic partner have been linked to changes in daily self-esteem levels for some individuals (Murray et al., 2003). Furthermore, more positive agentic narratives of a previous turning point were linked with more daily stability in evening self-esteem during a later life event (van Doeselaar & Reitz, 2022). Hence, believing that future experiences will turn out for the best may already boost self-esteem. We expected anticipation of social interactions to be positively related to evening self-esteem.

Trait self-esteem as moderator of the social interactions-self-esteem association (H3a, H3b)

Considerable individual differences have been found in daily self-esteem processes in previous studies (Denissen et al., 2008; Diwan et al., 2023). Therefore, we aimed to further explore the role of individual differences in the self-esteem-social interactions links. Specifically, we aimed to examine whether the magnitude of the short-term dynamics between momentary self-esteem and social interactions differed depending on trait self-esteem levels. Trait self-esteem is considered a relatively stable self-evaluation that is less responsive to particular situations than state self-esteem (Leary & Baumeister, 2000; Rubin & Hewstone, 1998).

Theoretical approaches pose conflicting ideas on how trait self-esteem might moderate the within-day associations between momentary self-esteem and social interactions. The sociometer theory suggests that individuals with low trait self-esteem react more strongly when real or imagined interpersonal difficulties arise than individuals with high trait self-esteem who already feel valued and included (Leary & Baumeister, 2000). Based on this reasoning, we anticipated that especially individuals low in trait self-esteem are reactive to social cues and have the strongest within-day associations between self-esteem and social interactions. Contrary, the self-verification theory poses a different idea on how trait self-esteem may moderate the link between momentary self-esteem and social interactions within days. Swann and Buhrmester (2012) proposed that individuals use their self-esteem to self-verify, for instance through creating their own social environments and interpreting situations in a self-confirmatory manner. This provides a sense of coherence and continuity, which is useful in guiding social interactions (Swann & Buhrmester, 2012). Following this reasoning, it is conceivable that individuals experience and anticipate social interactions in daily life that are in line with their trait self-esteem.

Previous research provided some empirical support for momentary associations between self-esteem and social interactions being moderated by trait self-esteem. Schmidt et al. (2020) conducted exploratory analyses and found that the within-day associations were weakest for children who had high average self-esteem. Wagner et al. (2023) studied the link between self-esteem reactivity to social inclusion and trait self-esteem development over time, which is a bottom-up approach instead of the top-down approach used in the present study. Their results showed no effect of self-esteem reactivity on the intercept of trait self-esteem, indicating that self-esteem reactivity did not predict mean trait self-esteem levels at the first measurement point. The effect of self-esteem reactivity on changes in trait self-esteem over time has not been examined due to a lack of interindividual variability in trait self-esteem change.

Moreover, several studies found that other measures underlying individual differences (i.e., personality, attachment styles) moderated the association between self-esteem and interpersonal measures. For instance, Murray et al. (2003b) found between-day associations for daily self-esteem and acceptance of romantic partner only for individuals who felt highly regarded by their partner. In addition, Srivastava and Beer (2005) found in a longitudinal study that while self-evaluations did not predict the extent to which individuals were liked by others on average, individual differences in attachment styles moderated the association. Specifically, they found a self-broadcasting effect for high anxious individuals, meaning that more negative self-evaluations were linked to being less liked by others, and a reversed self-broadcasting effect for low anxious individuals, meaning that lower self-evaluations were linked to being more liked by others. Finally, Bleckmann et al. (2022) found that momentary self-esteem reactivity to experiences of social inclusion was highest for individuals scoring high on agreeableness. Based on the theories and empirical research presented above, we expected a moderation effect of trait self-esteem on the within-day associations between self-esteem and quality/anticipation of social interactions. Given the conflicting ideas on how trait self-esteem may moderate the within-day associations, we had no a priori predictions about the direction of this moderation effect.

The present study

The present preregistered study examined the association between momentary self-esteem and quality of social interactions in daily life. We used data from four experience sampling measurement waves across one year, with each measurement burst consisting of four momentary assessments per day across 14 days. We had three research aims. First, we examined the within-day associations between morning/evening self-esteem and the quality of social interactions (measured as an aggregate score of three assessments during the day). Specifically, we expected that morning self-esteem predicted the quality of social interactions during the day (H1a), and that the quality of social interactions during the day positively predicted evening self-esteem (H1b). Second, we examined whether the positive/negative anticipation of social interactions (measured in the morning) was related to morning self-esteem (H2a) and evening self-esteem (H2b). Third, we estimated cross-level interactions to examine if the within-day associations between self-esteem and quality/anticipation of social interactions differed depending on trait self-esteem levels. We hypothesized a moderation effect for both the quality (H3a) and anticipation of social interactions (H3b), but had no a priori expectations regarding the direction of this effect.

Method

Procedure

The present study was part of a larger research project that investigated the development of self-esteem in young adults who transition from university to work. The research project was approved by the Ethical Review Board of Tilburg University. The codebook including the procedure and the questionnaires of this research project can be found on the Open Science Framework (https://osf.io/6nwjm). The present study was preregistered before data analysis (https://osf.io/4xenz).

Data were collected at Dutch universities from 2020 to 2021 across four measurement waves collected with four month intervals between each wave. Participants completed an online questionnaire at the start of every wave. Next, participants completed four short experience sampling questionnaires per day during a period of 14 days to assess daily experiences such as social interactions and momentary self-esteem. The experience sampling questionnaires were assessed via participants’ smartphones within fixed intervals at 6.00–11.00, 12.00–14.30, 16.30–19.00, and 21.00–04.00 using the Ethica Data Software (Ethica Data, 2021). To ensure high compliance in the experience sampling study we used small monetary rewards, monitored participants’ progress, and motivated participants to be compliant. Specifically, the compliance of participants was monitored and communicated via personalized messages informing on the percentage of completed ESM questionnaires and the amount of earned lottery tickets. The personalized messages also contained motivational messages about why their participation was valuable and important.

Participants

The first four waves of the data were used for the present study. A total of 232 students (M _age = 24.58 years, SD _age = 2.51 years, 69.9% female, 93.9% Dutch) started in the first wave, after excluding two participants who did not complete any ESM questionnaires. Participants were graduate students from Dutch universities in the final months before obtaining their Master’s (MSc) degree in different fields of study (economics, law, social sciences, humanities/digital sciences, theology). The total number of participants in each wave, the number of ESM questionnaires they completed, and the corresponding compliance rates are displayed in Table 1. In total, the data across all four measurement waves comprised of 23,107 data points. The compliance rates varied from 63.1 to 87 percent. We did not exclude participants based on missingness rates, but conducted sensitivity analyses to explore the robustness of our results against high missingness, by also analyzing data from only those participants that met certain compliance criteria.

Table 1.

Description of Experience Sampling Data.

		T1			T2			T3			T4
		N ppt	N obs	Compl.	N ppt	N obs	Compl.	N ppt	N obs	Compl.	N ppt	N obs	Compl.
Beep	6.00–11.00	230	2395	77.6	136	1372	69.1	117	1276	78.3	87	1053	87.0
	12.00–14.30	229	2266	73.5	135	1288	64.9	115	1207	74.2	87	924	76.5
	16.30–19.00	224	2236	73.0	136	1265	63.9	116	1176	72.4	85	889	73.8
	20.00–04.00	223	2334	76.3	133	1246	63.1	116	1193	73.7	85	987	82.1

Note. ppt. = participants; Obs = number of observations; Compl. = percentage of completed observations from total possible observations.

The present study made use of pre-existing data. A previous experience sampling study using the same dataset conducted Monte Carlo power simulations and found that a sample of 200 participants was sufficient to detect small prospective within-day associations (.07) between momentary self-esteem and a different momentary construct (pride) in a multilevel-VAR model with more than 80% power (Diwan et al., 2023). A more detailed explanation of these power simulations can be found here: https://osf.io/mfvp2.

Measures

Baseline questionnaires

Trait self-esteem

Trait self-esteem was measured with the Rosenberg Self-Esteem Scale (Rosenberg, 1965). This 10-item questionnaire used a five-point Likert scale instead of the original four-point scale. Answer options ranged from strongly disagree to strongly agree. Example items are: “On the whole, I am satisfied with myself,” “I feel I do not have much to be proud of,” and “I feel that I’m a person of worth, at least on an equal plane with others.” Cronbach’s alpha for this measure was .86.

Experience sampling questionnaires

Morning/evening self-esteem

Morning and evening self-esteem were measured with one item (“How satisfied are you with yourself right now?”) on a 5-point Likert-scale ranging from not at all to very much. This item has been used in a previous experience sampling method studies to assess self-esteem on a momentary scale (Valkenburg et al., 2021). Single-item measures of self-esteem have shown strong validity in previous research (Robins et al., 2001). Even though momentary self-esteem was assessed at all daily measurement occasions, we only included the measures assessed at the first (morning self-esteem) and the fourth measurement occasion (evening self-esteem) to capture the within-day associations before and after most social interactions took place.

Quality of social interactions

The quality of social interactions indicated the aggregated pleasantness score of all salient social events experienced during the day (after the first beep and before the last beep of the day). At the second, third, and fourth ESM assessment of the day, participants were asked to report on their most relevant personal events in the morning, noon, and evening (prior to the beep of the corresponding ESM questionnaire). The computation of this item involved multiple steps. First, participants reported how pleasant this specific event was (1: very unpleasant to 7: very pleasant). Next, they were asked to describe the kind of event by answering a multiple-choice question (a: interpersonal or social, b: health related, c: emotional or psychological, d: daily living, e: free time/sport/hobby, f: housing related, g: occupational/work-related, h: educational/school-related, i: other, j: none). Based on this item, we created a new binary variable in which we recoded all events that contained interpersonal/social aspects to one and the remaining events to zero. This binary item illustrated the presence/absence of interpersonal/social aspects in the events and was used as input for the computation of the quality of social interactions item. Specifically, when observations contained events that did not include interpersonal/social aspects, we coded a missing value for the quality of social interactions item. Observations that included interpersonal/social events were linked to the pleasantness scores. The final step for computing the quality of social interactions items during the day was to aggregate the pleasantness scores of all interpersonal/social events during the day. Simultaneously visualizing the social interactions of all measurement occasions within a day showed good internal consistency for the aggregated social interactions scores during the day (see Supplemental Material Figure 1).

Anticipation of social interactions

The anticipation of social interactions items indicated the extent to which participants looked forward (or not) to the day due to interpersonal reasons. These items were only collected in the first two measurement waves. We created two items for anticipation: one for positive anticipation (looked forward to the day due to interpersonal reasons) and one for negative anticipation (did not look forward to the day due to interpersonal reasons). For our research question we were not necessarily interested in the single effects of negative and positive anticipation, however, these separate variables were created because this was more convenient given the variables available in the secondary dataset. The positive and negative anticipation items were computed based on the following items respectively: “Due to what kind of reasons are you looking forward to the day (pick all that apply)?” and “For what reasons are you not looking forward to the day (pick all that apply)?”. These questions were assessed in the first ESM questionnaire of the day. We were only interested in observations in which participants anticipated interpersonal/social experiences and therefore selected the observations that contained interpersonal/social reasons for (not) looking forward to the day. We stored this information in two binary items: positive anticipation (looked forward to the day due to interpersonal reasons: yes/no) and negative anticipation (did not look forward to the day due to interpersonal reasons). Our hypothesis was supported if we found a positive association between anticipation and momentary self-esteem, meaning a positive association between positive anticipation and momentary self-esteem and a negative association between negative anticipation and momentary self-esteem.

Analytic strategy

Main analyses

Model specification

The hypotheses for the within-day associations (H1a, H1b, H2a, H2b) were tested with three multilevel (DSEM) models: one model for quality of social interactions, one model for negative anticipation of social interactions, and another model for positive anticipation of social interactions. These models included the within-day associations in both directions (morning self-esteem (IV) → (anticipated) social interactions (DV); (anticipated) social interactions (IV) → evening self-esteem (DV)). The associations were modeled at the within-level by regressing social interaction quality on morning self-esteem and regressing evening self-esteem on social interaction quality. All effects on the within-level (i.e., slopes, residual variances, autocorrelated residuals) were modeled as random to allow these parameters to vary across participants.

The stability of momentary self-esteem during the day was modeled by regressing evening self-esteem on morning self-esteem. We included the stability effect of momentary self-esteem to control for potential confounding effects of morning self-esteem when predicting evening self-esteem in the within-day association; we did not have hypotheses concerning the stability effect itself nor did we test mediation effects. In addition, we took possible autocorrelation between successive days into account by specifying first-order autoregressive structure for the residuals of morning self-esteem, evening self-esteem, social interaction quality, and positive/negative anticipation. The addition of autocorrelated residuals allowed us to account for variance in ESM items that was predicted by values at the previous day (Myin-Germeys & Kuppens, 2022). The stability effects and autocorrelated residuals for morning/evening self-esteem were separately modeled to disentangle and estimate different sources of variance. We expected that individuals experiencing high evening self-esteem may not only have higher morning self-esteem on the same day (possible stability effect), but might also experience higher evening self-esteem in general as explained by evening self-esteem values of previous days (possible autocorrelation effect).

To examine whether individuals differed in the within-day associations depending on their trait self-esteem levels (H3a, H3b), we added cross-level interactions to the multilevel models. The cross-level interactions were specified as a separate regression analysis on the between-level. Specifically, the random slopes of the within-day associations were the dependent variables and trait self-esteem was the independent variable. Each participant in our sample had a slope variable, and with the cross-level interactions we were able to examine if slopes for the within-day associations were different depending on individuals’ trait self-esteem levels.

Adding the cross-level interactions of the within-day associations for both morning and evening self-esteem in the same model might bias the results. To solely test the effect of each cross-level interaction, we added the cross-level interactions separately in six multilevel models. We tested multiple effects, but since it was a first investigation and given the ongoing discussion within the Bayesian field whether multiple testing is warranted (Harrell, 2023; Powell, 2019a, 2019b; Sjolander & Vansteelandt, 2019), we decided to not make the tests too stringent and therefore not correct for multiple testing. We instead replicated the models in different waves to increase the confidence in the results.

Model estimation

All models were estimated within the Dynamic Structural Equation Modeling (DSEM) framework in Mplus version 8 (Muthén & Muthén, 1998–2017). DSEM combines four modeling techniques: multilevel modeling, time-series modeling, structural equation modeling, and time-varying effects modeling (Asparouhov et al., 2017). Within this framework, the complex nested structure of the daily assessments is taken into account. DSEM uses a Bayesian estimator with Markov Chain Monte-Carlo algorithms. The standardized results provided in Mplus were used for model interpretation which allowed us to interpret the results in terms of effect sizes. Given the Bayesian approach of the DSEM analyses, we interpreted the parameter estimates using credible intervals. We did not have a priori beliefs about the priors, and therefore used the default in Mplus that assigns the parameters with uninformative priors (McNeish & Hamaker, 2020). Based on recommendations from previous studies (Brown et al., 2011; Hoofs et al., 2018), we set BCONVERGENCE to .005, the minimum number of iterations to 2000, and the Potential Scale Reduction (PSR) factor to <1.01. Models did not converge if the PSR was >1.01 after the maximum number of iterations was reached. In case of non-convergence, we simplified the models (i.e., by constraining random effects to be fixed).

To account for the binary scale of the positive anticipation item, we added a CATEGORICAL command in the corresponding scripts of these models. The TINTERVAL option in Mplus was used to account for unequal time intervals between ESM estimates, which can occur if participants do not report social events in successive measures, or in the case of missing observations. The TINTERVAL command specifies the data structure of the ESM data by defining the width between time lags, such that the number of time lags between the two available ESM estimates is known to the model (McNeish & Hamaker, 2020).

Model assumptions

As preregistered, we checked whether the models converged. The trace plots showed that convergence went well. In our preregistration we also mentioned that stationarity was an assumption for our analyses. However, this assumption was not relevant as we decided to correct for autocorrelation between closely sampled repeated measures with autocorrelated residuals instead of using autoregressive models.

Sensitivity analyses

We did not exclude observations or participants based upon careless responding or non-compliance since there was no consensus on data exclusion criteria yet. Some researchers suggest that the quality of ESM data can be affected by careless responding and non-compliance (Eisele et al., 2020; Schneider et al., 2018), while a simulation study showed that excluding non-compliant participants (≥20–30% missing) can actually bias the parameter estimates (Jacobson, 2020). Therefore, we flagged non-compliant participants and potential careless responders to explore the effects of careless responding and non-compliance on the data quality. The criteria for low within-person standard deviations (≤ .10), high within-person standard deviations (≥ .90), and low compliance (< 13.5%) were based on previous research (Meier et al., 2011; Schmidt et al., 2020). In the sensitivity analyses, we reran the analyses excluding flagged participants and compared these results to the previous results. If the interpretation of the results was different, we further explored whether the results were different due to potential careless responding, non-compliance, or both.

Deviations from preregistration

We encountered some unforeseen issues that resulted in some deviations from the preregistration, which are shortly discussed in this paragraph. First, we deviated from the analysis plan by not analyzing the data using autoregressive models but multilevel regression models within the DSEM framework, as these fitted our research questions better. Possible autocorrelation between successive days was taken into account by specifying a first-order autoregressive structure for the residuals. This better fitted our research questions, since our main interest were the within-day associations between self-esteem and social interactions and not the autoregression of our momentary measures. This decision simplified the model in terms of estimation and interpretation of results. Moreover, we left out the preregistered post-hoc power analyses based on the advice from a methodologist and the overall critique from methodologists regarding the utility of such analyses. According to Lakens (2022), post-hoc power analyses give biased results of power estimates because they assume that the effect size found in the study is the true effect size—which is not necessarily the case. Hence, such power estimates do not reflect actual power but rather present the results in a different way (Lakens, 2022). Instead, we replicated our analyses across three additional waves including 13,876 additional data points to examine the robustness of our results.

Results

The descriptive statistics of and intercorrelations between morning self-esteem, evening self-esteem, quality of social interactions, and positive anticipation are presented in Table 2. The quality of social interactions item correlated most with evening self-esteem (r range: 0.232–0.336), and to a smaller extent with morning self-esteem (r range: 0.091–0.183), trait self-esteem (r range: 0.084–0.265), and positive anticipation of social interactions (r range: 0.071–0.140). The positive anticipation of social interactions item correlated to both morning self-esteem (r range: 0.129–0.137) and evening self-esteem (r range: 0.090–0.140). These intercorrelations indicate the average effects across all participants and give an indication of the average associations between the momentary outcomes. In the main analyses, we also took individual differences in these associations into account by modeling random effects.

Table 2.

Descriptive Statistics and Intercorrelations of Main Outcomes.

	T1						T2						T3					T4
	M	SD	Intercorrelations				M	SD	Intercorrelations				M	SD	Intercorrelations			M	SD	Intercorrelations
			2	3	4	5			2	3	4	5			2	3	4			2	3	4
1. Morning self-esteem	3.12	.86	.526	.091	.321	.129	3.15	.84	.601	.170	.352	.137	3.15	.83	.601	.118	.369	3.05	.88	.630	.183	.400
2. Evening self-esteem	3.26	.91	—	.269	.271	.140	3.25	.90	—	.232	.375	.090	3.24	.89	—	.250	.334	3.23	.93	—	.336	.422
3. Quality SI	5.84	1.24	—	—	.084	.071	5.64	1.32	—	—	.140	.140	5.76	1.19	—	—	.167	5.74	1.32	—	—	.265
4. TSE	3.71	.59	—	—	—	.050	3.78	.57	—	—	—	−.012	3.73	.57	—	—	—	3.71	.62	—	—	—
5. Pos. antic.	.42	.49	—	—	—	—	.35	.48	—	—	—	—	—	—	—	—	—	—	—	—	—	—

Note. M = mean; SD = standard deviation; Quality SI = Quality of experienced social interactions during the day; Pos. antic. = Positive anticipation of social interactions.

The mean and median number of social interactions reported per participant across the measurement waves T1, T2, T3, and T4 were, respectively: M = 10.7, 8.6, 9.1, 10.5; Mdn = 10, 8, 9, 11. Positive and negative anticipation of social interactions were only assessed in measurement waves T1 and T2. Across these two measurement waves, the mean and median number of observations were respectively: M = 5.6, 4.8; medians = 5 for positive anticipation, and M = 0.43, 0.55; medians = 0 for negative anticipation. The distribution of the total number of reported social interactions per participant can be found in Table S5 of the Supplemental Materials. The distribution of the total number of observations including positive or negative anticipation per participant can be found in Tables S6 and S7 of the Supplemental Materials.

The intraclass correlations were computed for each measurement wave to assess how much variation in our momentary variables was due to within- and between-level variance. The intraclass correlations showed that the variation in morning/evening self-esteem was explained half on the between-person and half on the within-person level. Variation in quality of social interactions during the day and positive anticipation of social interactions were mostly explained on the within-person level. For negative anticipation we found that almost all variation was explained on the between-person level with no within-person variance for most participants. The results of the within-day associations and cross-level interactions are displayed in Tables 3–7 and summarized in Table S1 of the Supplemental Materials.

Table 3.

Within-Day Associations Between Morning/Evening Self-Esteem and Quality of Social Interactions Across Four Waves.

	T1			T2			T3			T4
	Estimate	95% CI	SD	Estimate	95% CI	SD	Estimate	95% CI	SD	Estimate	95% CI	SD
Iterations	38,200	—	—	24,000	—	—	16,600	—	—	14,400	—	—
N Obs. (N ppt)	2,876 (231)	—	—	1,727 (139)	—	—	1,510 (119)	—	—	1,169 (87)	—	—
Morning SE → evening SE	.183 a	[.129, .237]	.028	.152 a	[.083, .221]	.036	.154 a	[.069, .225]	.040	.166 a	[.068, .257]	.048
Morning SE → quality social interactions	.043	[-.010, .089]	.026	.050	[-.057, .135]	.048	.038	[-.045, .124]	.041	−.052	[-.162, .052]	.055
Quality social interactions → evening SE	.150 a	[.104, .200]	.025	.162 a	[.086, .245]	.042	.160 a	[.097, .230]	.034	.205 a	[.122, .306]	.047

95% CI = 95% credible interval; SD = posterior standard deviation; Ppt = participants; Obs. = observations; SE = self-esteem.

^aindicates that 0 is not in the Bayesian credible interval.

Table 4.

Within-Day Associations Between Morning/Evening Self-Esteem and Positive Anticipation of Social Interactions Across Two Waves.

	T1			T2
	Estimate	95% CI	SD	Estimate	95% CI	SD
Iterations	12,900	—	—	21,800	—	—
N Obs. (N ppt)	3,079 (231)	—	—	1,965 (147)	—	—
Morning SE → evening SE	.189 a	[.136, .243]	.028	.149 a	[.074, .224]	.038
Morning SE → positive anticipation	.148 a	[.065, .242]	.040	.111 a	[.023, .205]	.047
Positive anticipation → evening SE	.013	[-.010, .047]	.015	.012	[-.036, .058]	.024

95% CI = 95% credible interval; SD = posterior standard deviation; TSE = trait self-esteem; SE = self-esteem; Obs. = observations.

^aIndicates that 0 is not in the Bayesian credible interval.

Table 5.

Cross-level Interaction of Trait Self-Esteem on the Association Between Morning Self-Esteem and Quality of Social Interactions Across Four Waves.

	T1			T2			T3			T4
	Estimate	95% CI	SD	Estimate	95% CI	SD	Estimate	95% CI	SD	Estimate	95% CI	SD
Iterations	5,400	—	—	8,500	—	—	10,000	—	—	8,000	—	—
N Obs. (N ppt)	2,833 (227)	—	—	1,536 (123)	—	—	1,385 (109)	—	—	1,088 (81)	—	—
Stability momentary SE	.156 a	[.108, .200]	.024	.176 a	[.103, .249]	.038	.118 a	[.040, .191]	.039	.121 a	[.048, .198]	.038
Morning SE → quality social interactions	.043	[-.005, .097]	.026	.031	[-.083, .107]	.049	.049	[-.038, .151]	.048	−.017	[-.102, .065]	.042
Quality social interactions → evening SE	.193 a	[.143, .245]	.026	.146 a	[.033, .223]	.049	.248 a	[.157, .327]	.042	.271 a	[.182, .360]	.045
TSE → quality social interactions	.306 a	[.151, .462]	.080	.243 a	[.049, .434]	.099	.224 a	[.031, .413]	.097	.478 a	[.256, .688]	.110
TSE → within-day association between morning SE and quality social interactions	.073	[-.343, .452]	.080	−.016	[-.556, .848]	.335	.119	[-.431, .727]	.273	−.114	[-.433, .195]	.157

^aIndicates that 0 is not in the Bayesian credible interval. 95% CI = 95% credible interval; SD = posterior standard deviation; TSE = trait self-esteem; SE = self-esteem; Obs. = observations.

Table 6.

Cross-level Interaction of Trait Self-Esteem on the Association Between Quality of Social Interactions and Evening Self-Esteem Across Four Waves.

	T1			T2			T3			T4
	Estimate	95% CI	SD	Estimate	95% CI	SD	Estimate	95% CI	SD	Estimate	95% CI	SD
Iterations	6,400	—	—	15,300	—	—	10,200	—	—	9,100	—	—
N Obs. (N ppt)	2,833 (227)	—	—	1,536 (123)	—	—	1,385 (109)	—	—	1,088 (81)	—	—
Stability momentary SE	.154 a	[.112, .199]	.022	.190 a	[.114, .263]	.037	.114 a	[.035, .191]	.039	.118 a	[.044, .190]	.037
Morning SE → quality social interactions	.052	[-.003, .103]	.027	.015	[-.085, .122]	.053	.053	[-.023, .176]	.049	−.018	[-.099, .073]	.044
Quality social interactions → evening SE	.198 a	[.145, .252]	.027	.137 a	[.040, .217]	.044	.250 a	[.166, .326]	.040	.269 a	[.184, .348]	.042
TSE → evening SE	.020	[-.040, .082]	.032	.081 a	[.001, .163]	.041	−.032	[-.098, .047]	.037	.061	[-.026, .158]	.047
TSE → within-day association between quality social interactions and evening SE	.147	[-.060, .342]	.102	.048	[-.309, .400]	.180	.065	[-.159, .277]	.110	−.110	[-.455, .195]	.167

^aIndicates that 0 is not in the Bayesian credible interval. 95% CI = 95% credible interval; SD = posterior standard deviation; TSE = trait self-esteem; SE = self-esteem; Obs. = observations.

Table 7.

Cross-Level Interactions of Trait Self-Esteem on Association Between Morning/Evening Self-Esteem and Positive Anticipation Across Two Waves.

	T1			T2
Cross-level Interaction Morning Self-Esteem	Estimate	95% CI	SD	Estimate	95% CI	SD
Iterations	12,000	—	—	8,200	—	—
N Obs. (N ppt)	3,036 (230)	—	—	1,724 (128)	—	—
Stability momentary SE	.178 a	[.129, .223]	.024	.168 a	[.101, .230]	.033
Morning SE → positive anticipation	.148 a	[.087, .208]	.030	.093	[-.001, .175]	.044
Positive anticipation → evening SE	.002	[-.033, .053]	.020	.000	[-.086, .094]	.046
TSE → positive anticipation	.041	[-.068, .152]	.056	.001	[-.142, .139]	.071
TSE → within-day association between morning SE and positive anticipation	.206	[-.097, .695]	.193	.341	[-.202, .948]	.311
Cross-level Interaction Evening Self-Esteem	Estimate	95% CI	SD	Estimate	95% CI	SD
Iterations	16,800	—	—	10,000	—	—
N Obs. (N ppt)	3,036 (230)	—	—	1,724 (128)	—	—
Stability momentary SE	.167 a	[.110, .220]	.028	.179 a	[.109, .246]	.034
Morning SE → positive anticipation	.159a	[.090, .225]	.035	.094	[-.014, .172]	.046
Positive anticipation → evening SE	.013	[-.035, .063]	.024	−.007	[-.097, .078]	.046
TSE → evening SE	−.001	[-.055, .057]	.029	.074 a	[.002, .161]	.040
TSE → within-day association between positive anticipation and evening SE	.111	[-.697, .703]	.334	.139	[-.068, .333]	.101

^aIndicates that 0 is not in the Bayesian credible interval. 95% CI = 95% credible interval; SD = posterior standard deviation; TSE = trait self-esteem; SE = self-esteem; Obs. = observations.

Both stability effects and autocorrelated residuals predicted part of the variance in the ESM outcome measures. Within days, a positive stability effect from morning to evening self-esteem was found across all measurement waves, indicating that morning self-esteem partly predicted evening self-esteem within days. Across days, autocorrelations between residuals were found for all ESM outcomes, indicating that ESM estimates were partly predicted by their values on previous days. The autocorrelated residuals were highest for positive anticipation of social interactions.

Quality of social interactions and morning/evening self-esteem (H1a, H1b)

To examine whether momentary self-esteem was associated with quality of social interactions, we linked morning and evening self-esteem to the quality of experienced social interactions during the day. The standardized model results across all four measurement waves are displayed in Table 3. Across all four waves, morning self-esteem did not predict the quality of experienced social interactions during the day. Vice versa, we found that the quality of experienced social interactions during the day predicted evening self-esteem across all four waves. These results provided conclusive support for the hypothesis that the quality of social interactions during the day are positively related to evening self-esteem (H1b), whereas they are not in line with our hypothesis that morning self-esteem is related to quality of social interactions during the day (H1a). As a sensitivity analysis, we reran the analyses controlling for gender (Table S4), and found similar results as in the main analysis, suggesting that gender did not influence the results.

Anticipation of social interactions and momentary self-esteem (H2a, H2b)

To examine the within-day associations between momentary self-esteem and anticipation of social interactions, we associated morning and evening self-esteem to negative and positive anticipation. The standardized model results from the first two measurement waves are displayed in Table 4. Descriptive statistics for negative anticipation (Table S7) showed that in the first and second measurement wave, participants rarely experienced a negative anticipation of social interactions. For instance, in the first measurement wave, only 71 participants (30.6%) had a negative anticipation of social interactions in at least one of the 14 observations, with a total 100 observations in the sample (1.1%) in which negative anticipation of social interactions was experienced. Given this low prevalence and the potentially limited utility of these data to describe general patterns in the broader population, we only report the results for negative anticipation in the Supplementary Material (Table S2) and do not interpret the results here.

The positive anticipation models did not converge in the first instance, and therefore we had to constrain the autocorrelated residuals to be fixed to decrease the complexity of these models. After simplifying the positive anticipation model, results showed that morning self-esteem predicted positive anticipation of social interactions, while positive anticipation did not predict evening self-esteem across the two measurement waves. These results support our hypothesis that morning self-esteem positively relates to the anticipation of social interactions (H2a), but are not in line with the hypothesis that anticipated social interactions positively relate to evening self-esteem (H2b). As a sensitivity analysis, we reran the analyses controlling for gender (Table S3), and found similar results as in the main analysis, suggesting that gender did not influence the results.

Trait self-esteem as the moderator of within-day associations (H3a, H3b)

Person-specific effects sizes were estimated to explore the heterogeneity in the within-day associations. These person-specific effect sizes comprise the unique estimates for the proposed within-day associations of each participant in our sample based on their personal time-series data. Individuals differed considerably in how strong the within-day associations unfolded in their daily life (Figures S2–S5 depict the histograms for person-specific effect sizes).

Cross-level interactions: Trait self-esteem and quality of social interactions (H3a)

Cross-level interactions were added to the models to test whether trait self-esteem moderated the associations between the quality of social interactions and morning/evening self-esteem. The standardized model results are presented in Tables 5 and 6. We did not find cross-level interactions for trait self-esteem on the within-day associations between the quality of social interactions and morning/evening self-esteem across the four measurement waves. The results provided no support for the hypothesis that trait self-esteem moderates the within-day associations between momentary self-esteem and quality of social interactions (H3a). It should be noted that the credible intervals for the cross-level interactions (see Tables 5 and 6) were large in most models, indicating that the cross-level estimates for these models were not very certain given the current sample size. For instance, in the second measurement wave, there was a 95% probability that the true value for the standardized cross-level interaction was between −0.556 and 0.848. In comparison, the credible interval for the within-day association was much smaller (between −0.083 and 0.107), suggesting that the estimates for the within-day associations were more certain given the current sample size.

Cross-level interactions: Trait self-esteem and anticipation of social interactions (H3b)

Next, we tested cross-level interactions for trait self-esteem moderating the within-day associations between positive anticipation of social interactions and morning/evening self-esteem. Based on the credible intervals, we did not find a cross-level interaction of trait self-esteem on the within-day associations between positive anticipation and morning/evening self-esteem across the two measurement waves. These results are not in line with our hypothesis that trait self-esteem moderates the within-day associations between momentary self-esteem and the anticipation of social interactions (H3b). As before, the credible intervals for the cross-level interactions were large, indicating that the estimates were not very certain given the current sample size. For instance, there was a 95% probability that the true value of the standardized cross-level interaction was between −0.097 and 0.695 in the first measurement wave.

Sensitivity analyses

Our results suggested that almost half of the sample in the first measurement wave met at least one of the prespecified criteria for exclusion in sensitivity analyses (low within-person standard-deviations (≤.10, N = 34), high standard within-person deviations of (≥.90, N = 47), and/or low compliance (<13.5%, N = 23)). Excluding these participants did not change the results substantially. The models excluding participants with high within-person standard deviations did not converge. One potential explanation for non-convergence may be that the sample without these individuals did not contain enough variance. Another potential explanation is the loss of power when excluding these participants, since around one fifth of our sample was excluded when following our prespecified criteria. Even though we based our preregistered exclusion criteria for sensitivity analyses on previous research, for future research it is recommended to specify more strict criteria for sensitivity analyses. Finally, based on warning messages in Mplus, we also explored if excluding individuals with no within-person variance on more than two of the ESM outcomes changed the results, which was not the case.

Discussion

The present experience sampling study examined the within-day associations between quality/anticipation of social interactions and morning/evening self-esteem and explored the moderating role of trait self-esteem in these associations across one year including four ESM measurement waves. Our results add novel insight into the dynamic processes of momentary self-esteem and different aspects of social interactions (anticipation and quality) within days. Specifically, we found that morning self-esteem predicted a positive anticipation of social interactions, while evening self-esteem was predicted by the quality of social interactions experienced during the day. Moreover, we found individual differences in these within-day associations, but no support for trait self-esteem predicting these individual differences.

Morning/evening self-esteem and quality of social interactions during the day

A prospective association between social relationships and self-esteem has been consistently found in longitudinal studies (Harris & Orth, 2020), however, inconclusive support existed for the momentary association between self-esteem and social interactions in daily life. Most of the previous daily diary studies found that the quality of social interactions predicts momentary self-esteem across days, but not vice versa (Denissen et al., 2008; Schmidt et al., 2020). In line with these empirical studies, we found support for the quality of social interactions during the day predicting evening self-esteem across four measurement waves. This finding aligns with propositions from different theories suggesting that social relationships can shape self-esteem over time (Ainsworth, 1973; Bowlby, 1973; Cooley, 1902; Harris & Orth, 2020; Harter, 1999; Leary & Baumeister, 2000; Shrauger & Schoeneman, 1979). For future research it would be interesting to further examine the carry-over effects of daily social interactions on momentary self-esteem, for instance by adding an extra assessment of momentary self-esteem with a longer interval after the last assessment of social interactions.

Our results suggest a unidirectional instead of a bidirectional association between the quality of social interactions and momentary self-esteem within days. While we found a prospective effect of social interaction quality during the day on evening self-esteem, we did not find the reverse prospective effect during the day. This finding contradicts the longitudinal evidence of self-esteem and social relationships being prospectively associated in both directions across months and years (de Moor et al., 2021; Harris & Orth, 2020; Hutteman et al., 2015), but is in line with previous daily diary research zooming in on the short-term dynamics of the association (Denissen et al., 2008; Subrahmanyam et al., 2020)

This might indicate that self-esteem is prospectively related in both directions with social relationships on the long-term (e.g., months, years), but not in daily manifestations of social relationships (i.e., social interaction quality). This would be an interesting finding, as quite some studies found concurrent (in the moment) associations between momentary/state self-esteem and interpersonal measures in daily diary studies (Bleckmann et al., 2022; Denissen et al., 2008; Heppner et al., 2008; Mahadevan et al., 2020; Nezlek, 2005; Pietromonaco & Barrett, 2009; Schmidt et al., 2020; Subrahmanyam et al., 2020; Wagner et al., 2023). Possibly, the prospective association unfolds in a different timescale than the one used in the present study. In our study, we found that that the quality of the combined social interactions during the day was a strong predictor for evening self-esteem, but that morning self-esteem did not predict this aggregated social interaction quality score. It is possible that the effects of morning self-esteem are more short-lived and therefore only predict situation-specific measures of social interactions (no aggregate score) assessed closely related in time.

Morning/evening self-esteem and anticipation of social interactions

We expected that anticipatory thoughts about social interactions also play a role in the within-day associations between momentary self-esteem and social interactions. Prominent theories such as the sociometer theory suggest that self-esteem is not only reactive to actual but also imagined interpersonal threats (Leary & Baumeister, 2000). Likewise, empirical evidence shows that imagining social threats can already initiate state self-esteem changes (Leary et al., 1998). Hence, we examined if morning self-esteem predicted anticipation of social interactions, and if anticipation of social interactions predicted evening self-esteem.

Prevalence of negative anticipation: Generalizable or sample/measure characteristics?

The prevalence of observations including a negative anticipation of social interactions was low in our sample. Most participants rarely experienced a negative anticipation of social interactions. Given this low prevalence, we did not interpret the results for negative anticipation. Future studies are needed to determine whether the low prevalence of negative anticipation is generalizable to broader populations or rather due to the sample characteristics, measures, or timing of data collection. It may simply be that negative anticipation is not often experienced in daily life, which is in line with other experience sampling studies showing that participants less often report negative experiences in daily life (i.e., negative affect and social exclusion) than positive experiences (i.e., positive affect and social inclusion) (Bulow et al., 2022; Schmidt et al., 2020). There are also alternative explanations for the low prevalence of negative anticipation. The number of observations with negative anticipation might have been low because of sample characteristics, as the participants had above average trait self-esteem levels. This might have made them less likely to report negative anticipations of social interactions. Another possibility is that the binary measures of anticipation were not sensitive enough to detect more subtle effects of slightly negative anticipation. Using a continuous item of anticipation may result in a higher prevalence of negative anticipation since it can better detect these subtle effects.

Furthermore, the data collection took place during the Covid-19 pandemic, which might have limited opportunities for social interactions, especially those some people might dread (e.g., in-person work meetings that require higher social engagement than online meetings). It is likely that not only daily negative events, but also positive events were less frequently experienced or canceled during the pandemic. Assuming that negative anticipation is to a lesser extent experienced than positive anticipation in daily life, this low prevalence might have become even more noticeable in the presence of less opportunities for social interactions, and might explain the low number of observations with negative anticipation.

Morning self-esteem predicting positive anticipation of social interactions

The results from the present study showed that while positive anticipation in the morning was concurrently related to morning self-esteem, it did not predict evening self-esteem. These findings could be explained by different processes. First, it is possible that morning self-esteem responded to anticipating daily social events in the moment, but that these effects did not prevail until the end of the day. The end-of-the-day self-esteem seems instead rather responsive to experienced social interactions during the day. However, that anticipation might have a short-term effect on self-esteem is plausible and in line with studies showing that self-esteem can change in anticipation of major life events (Bleidorn & Schwaba, 2018; Reitz et al., 2022; van Scheppingen et al., 2018). A second possible explanation is the reverse effect: the anticipation of social events at the start of the day may influence the evaluation of self-esteem. For instance, looking forward to a daily social event in which one will engage with people they like might generate positive feelings that in turn make them feel better about themselves. Lastly, it is also possible that the association only unfolds concurrently and has no causal effect at all. In this situation, other daily processes may explain why these two constructs are related. For example, experiencing positive emotions or a positive outlook at the start of the day might result in positive outcomes on other momentary constructs as well. Feeling good in general might then explain why one feels better about themselves, and also looks forward to the daily social events during the day.

These different explanations of anticipation mechanisms warrant further investigation as they have important implications for theory and practice. If future research corroborates a directional effect, it would inform self-esteem theory in that not only experiencing but also anticipating social interactions might play a role in daily self-esteem dynamics. Better insights can help answer the question of what comes first: Is self-esteem in the moment determining our perceptions of anticipated social interactions (concurrently or short-lived)—or is it rather that our perceptions of anticipated social interactions determine how we feel about ourselves? These insights into possible anticipation mechanisms can also have important implications for clinical practice. Namely, it might be that intervening on momentary self-esteem or positive emotions at the start of the day is a first promising step in breaking the vicious cycle between low self-esteem and the lack of high-quality social interactions.

Moreover, it would be very interesting to examine if these anticipation effects are differential across interaction partners. For instance, it is conceivable that the anticipation effects are different for “self-chosen” social interaction partners (i.e., friends, romantic partner) compared to occasional social interaction partners (i.e., colleagues, acquaintances). And if differences are found, it is worth investigating if these differences are due to the closer/more supportive ties with self-chosen interaction partners compared to occasional social partners, or if there are any other mechanisms at place. A better understanding of anticipation effects for different interaction partners can provide insights into the most effective way to break the vicious circle between low self-esteem and low quality social interactions.

No support for positive anticipation predicting evening self-esteem

We did not find an effect of positive anticipation predicting evening self-esteem, even though we expected an effect based on propositions from theory and preliminary empirical evidence (Leary & Baumeister, 2000; Leary et al., 1998). One possible explanation for this null finding is that positive anticipation effects only occur concurrently and do not have long-lasting effects within days. This in line with the finding that positive anticipation predicted morning self-esteem within the first ESM assessment of the day, but did not predict evening self-esteem hours later. Leary et al. (1998) argued that state self-esteem is more reactive to rejection than approval, so it is conceivable that positive anticipation does not evoke strong effects on momentary self-esteem across wide time intervals. It might still be that positive anticipation has carry-over effects on momentary self-esteem across shorter time scales then assessed in our study design (i.e., within a few hours instead of a day). More research using different timescales is needed to better understand the possible carry-over effects of positive anticipation on momentary self-esteem within days.

Trait self-esteem: No support for a moderating role

The magnitude of the within-day processes differed considerably across individuals, which is in line with previous research showing that daily self-esteem predicts the quality of social interactions across days for some individuals, while this is hardly the case for other individuals (Denissen et al., 2008). The heterogeneity in the person-specific effect sizes of the within-day associations highlight that the associations between morning/evening self-esteem and quality/anticipation of social interactions do not similarly unfold in daily life for different individuals. No one size fits all when it comes to these associations, hence, looking into certain individual characteristics might help us better understand these individual differences.

Building on propositions from different theories about social interactions and self-esteem (Leary & Baumeister, 2000; Murray et al., 2000; Srivastava & Beer, 2005; Swann & Buhrmester, 2012), we expected that momentary self-esteem can have different functions for different individuals, that is, to monitor one’s relational value, to self-verify in social interactions, to elicit behaviors from others, or to have no function at all. We therefore explored the role of individual differences in these daily self-esteem processes and hypothesized that trait self-esteem moderates the within-day associations between social interaction quality/anticipation and morning/evening self-esteem. A better understanding of individual differences in self-esteem reactivity to social interactions in daily life is important as it can guide interventions that promote momentary self-esteem (Wagner et al., 2023). In line with these expectations, previous daily diary studies found that the associations between momentary self-esteem and social outcomes differ across different individuals, for example, for individuals who feel highly regarded by their partner (Murray et al., 2003b), have a low or high anxious attachment style, score high on agreeableness (Bleckmann et al., 2022), or when a male was interacting with a close friend (Denissen et al., 2008).

When adding trait self-esteem to the within-day associations with a cross-level interaction, we found no support for the hypothesis that trait self-esteem moderates the within-day associations. A possible explanation is that our sample mostly contained high trait self-esteem individuals who mostly reported high quality social interactions. This variance restriction in trait self-esteem may come with limited variance in within-day associations across individuals. Replication studies using samples with more diverse levels of trait self-esteem are needed to examine whether there is more variability in within-day associations depending on trait self-esteem levels. Another possible explanation is that the sample was not large enough to detect cross-level interactions. Replication studies using larger samples are needed to examine if our null findings are generalizable to other samples or rather due to a lack of power.

Limitations and future directions

The present study had several strengths. It is one of the first studies to test the within-day associations between momentary self-esteem and different characteristics of social interactions (quality and anticipation) across four waves of experience sampling data. Our study not only assessed the inter-individual variability in these associations but also explored to what extent this variability can be explained by trait self-esteem. Our hypotheses and analytic plan were preregistered before data analysis. The study also had some limitations.

First, the experience sampling method used to assess prospective links between momentary self-esteem and social interactions in daily life is still in its infancy. The lack of knowledge about effect sizes from previous research and the complexity of the experience sampling data and multilevel analyses with many parameters complicated a priori power analyses for our specific research questions. We expected to have enough power for the within-day associations based on the power simulations of a previous study using the same dataset, showing sufficient power to detect small prospective within-day associations between self-esteem and positive affect (Diwan et al., 2023). For the cross-level interactions, it remains unclear if we had sufficient power to detect cross-level effects, as we were unable to test the power due to a lack of a priori information that is required for such analyses (Mathieu et al., 2012). Therefore, we cannot exclude the possibility that the null findings for the cross-level interactions of trait self-esteem were due to the lack of power. Future research using larger samples is needed to replicate the moderation effect.

Moreover, even though the single-item measures of momentary self-esteem showed sufficient convergent validity, we were unable to examine the reliability of this item. For future research it is recommended to use a multiple-item version of momentary self-esteem and see if our results replicate, which we do expect given that single-item measures of momentary constructs (Cloos et al., 2022) and self-esteem (Robins et al., 2001) have shown to be efficient low-burden assessment tools.

Another limitation is the potential limited generalizability of our sample that mostly consisted of females followed from a few months before until a year after their graduation from a Master’s program. Future research is needed to determine if our results are generalizable to other populations, although there is some preliminary evidence suggesting that effects in a graduate sample also hold for the broader population, as Wagner et al. (2023) found similar within-day dynamics of momentary self-esteem and quality of social interactions for graduates and pensioners.

Conclusion

The present study demonstrates that differentiating between anticipated and experienced social interactions provides a more fine-grained understanding of the dynamic processes between self-esteem and social interactions. Our results not only showed that social interaction quality during the day predicted evening self-esteem, but we also found that fluctuations in self-esteem co-occur with the anticipation of positive daily social events at the start the day. Interestingly, we did not find anticipation effects across wider time intervals (no link with evening self-esteem), and also no associations between morning self-esteem and social interaction quality during the day. These findings may start a debate on how to best study daily self-esteem processes in relation to social interactions in experience sampling studies, both in terms of measuring social interactions (our study shows that both experienced quality and anticipation of social interactions are important) and choosing the best time intervals. There are currently no clear guidelines on how to best capture social interaction-self-esteem links in daily life, and it remains uncertain if our null findings indeed indicate non-existing effects or rather that effects are more short-lived and only exist within shorter time scales. A better understanding of these dynamics in daily life is crucial for clinical practice as well, as they might inform daily interventions (i.e., boosting momentary self-esteem or positive emotions) that can help break vicious circles between low self-esteem and low quality social interactions/relationships.

Overall, individuals differed substantially in how the association between anticipated/actual social interactions and momentary self-esteem unfolded in their daily life. We found no support for the hypothesis that trait self-esteem moderates the within-day associations between momentary self-esteem and the anticipation/quality of social interactions within days. As no one size fitted all for these daily self-esteem-social interactions links, it is important that more research on individual differences in daily self-esteem-social interactions links is conducted, to examine if our results replicate in a larger and more diverse sample and to explore if any other personality traits are at play.

Supplemental Material

Supplemental Material - Self-esteem and social interactions in daily life: An experience sampling study

Supplemental Material for Self-esteem and social interactions in daily life: An experience sampling study by Manon Enting, Joran Jongerling, and Anne Kristin Reitz in European Journal of Personality