Understanding the relationship between Facebook use and adaptation to financial hardship: Evidence from a longitudinal panel study.

Prior longitudinal studies on the effects of Facebook use on well-being often found no or only small effects. One reason could be that well-being indicators are often remarkably stable over time. In the present study, we looked therefore at people who experienced financial hardship, a stressful life event, and examined whether Facebook users differed from non-users in how they reacted and adapted to the life event over time and which role social support played in this process. We used multilevel models to examine the recovery process to the negative life event. Facebook users experienced a larger drop of satisfaction with their financial situation during the life event (= the reaction phase) than non-users, but showed higher levels of satisfaction after the life-event (= the adaptation phase). Online social support was also beneficial for adaptation to the life event. Next, we examined within the subsample of Facebook users to what extent frequency of posting and reading were related to receiving online social support and in turn financial satisfaction. Frequency of posting and reading were positively related to online social support. However, only posting was related to higher financial satisfaction and only at the end of the adaptation phase, indicating that Facebook use might mainly contribute positively to people's well-being some while after stressful life-events.

Introduction

Facebook use has become part of the daily routine for many people, and researchers have wondered how Facebook use affects life-satisfaction or other indicators of well-being. A recent review (Verduyn, Ybarra, Résibois, Jonides, & Kross, 2017) showed that active Facebook use, i.e. posting status updates and writing messages, is good for well-being, mainly because people build social capital from which they can receive social support. Passively browsing the mainly positive updates of others, in contrast, negatively affects well-being because the resulting social comparisons trigger envy. The majority of studies has however been cross-sectional or used student samples (Verduyn et al., 2017). Longitudinal analyses found a less clear picture. Effects occurred either only for one of the well-being indicators (Dienlin, Masur, & Trepte, 2017), the reverse paths were also significant (Reinecke & Trepte, 2014), or effects occurred only for very specific forms of communication (e.g., targeted communication with strong ties; Burke & Kraut, 2016). Most longitudinal studies contained only two waves (Dienlin et al., 2017, Reinecke and Trepte, 2014) or did not analyze the temporal patterns (three waves, Burke & Kraut, 2016; four waves, Trepte, Dienlin, & Reinecke, 2015), making it difficult to detect non-linear trajectories.

A recent study by Utz & Breuer (2017) analyzed six waves of a panel study with a representative sample of Dutch internet users and found that Facebook users reported higher levels of online social support, but also higher levels of stress. Users and non-users did not differ in life satisfaction. Detailed cross-lagged panel analyses within the group of users showed that Facebook use, more specifically, asking for advice, and receiving social support online were positively related within as well as across waves. Higher levels of online social support, however, did not result in a decrease of stress or an increase of life satisfaction in subsequent waves (see also Trepte et al., 2015). Instead, stress and online social support were positively correlated within waves. This could indicate that a third variable, such as a stressful life event, drives both, higher levels of stress and higher online social support.

To explore this puzzling finding further, we focus in the present paper only on people who experienced a stressful life event. We chose the experience of financial hardship for pragmatic and content-related reasons. The pragmatic reason is that experiencing financial hardship was one of the most frequently encountered life events in our sample, ensuring a larger subsample and thus higher power. Content-wise, in contrast to more ambiguous life events that might be positive for some people (e.g. moving, relationship break), financial hardship is clearly negative; and it has been shown that it is related to depression (Butterworth, Rodgers, & Windsor, 2009).

The goal of the present paper is to examine the role of Facebook use in recovering from life events such as financial hardship. It examines whether Facebook users and non-users differ in how experiencing financial hardship affects their satisfaction with financial aspects of their life before, during and after the life event. We zoom in into the role of online social support in recovering from experiencing financial hardship. Finally, we examine within the subsample of Facebook users the role of active and passive Facebook use (Burke, Kraut, & Marlow, 2011). Our work not only sheds more light on the role Facebook use can play for well-being, but it also contributes to research on adaptation to life events. Recent work has shown that different people react differently to life events, but this work has so far focused on personality characteristics (Hirsch et al., 2007, Yap et al., 2012) and neglected the role of media use.

Theoretical background

Research on subjective well-being, i.e., how people feel and think about their lives, distinguished between an affective and a cognitive component (Diener, 1984). Whereas affective well-being refers to the presence/absence of pleasant/unpleasant affect, cognitive well-being refers to the cognitive evaluation of life. The cognitive evaluation, i.e. the judgment of outcomes as fulfilling expectations, can be measured on a global level, but also for certain domains such as career satisfaction (Luhmann, Hofmann, Eid, & Lucas, 2012). A recent meta-analysis on the effects of various life events on well-being has shown that effects are usually larger on cognitive well-being (Luhmann et al., 2012). The present paper therefore focuses on cognitive well-being, more specifically, satisfaction with one's financial situation (or shorter: financial satisfaction).

Adaptation to life events and life satisfaction: the set-point model

Research on the effect of life events on well-being has long been dominated by the assumption that people can and do adapt to life events (for an overview, see Lucas, 2007). The term hedonic treadmill has been coined by Brickmann and Campbell (1971) to describe the fact that people temporarily react on life events, but that they adapt soon back to their personal set-point of life satisfaction. Due to this quick adaptation and habituation to new circumstances, life satisfaction often seems remarkably stable in longitudinal studies. This view has been challenged, as the review by Diener, Lucas, and Scollon (2006) shows. It has for example been shown that people have individual set-points, also influenced by personality, and that they have multiple set-points for different domains, e.g. life in general vs. work. More important, people differ in their ability to recover from negative life events. Studies using panel data, following a large number of people over a long time, have for example found that not everybody recovers from widowhood (Lucas, Clark, Georgellis, & Diener, 2003).

In longitudinal panel studies taking a process perspective, life satisfaction before the life event is usually treated as baseline and compared with life satisfaction during the life event (the reaction phase) and life satisfaction in the years after the life event (the adaptation phase). The variance in trajectories is then studied. Groups of people could differ for example in the size of the drop in life satisfaction in the reaction phase or in whether their life satisfaction recovers until the baseline level of stays at a lower level in the adaptation phase. Differences in adaptation patterns have usually been explained with interindividual differences in coping strategies or personality characteristics (Diener et al., 2006, Hirsch et al., 2007). The effects of media use have been neglected in this research tradition, and social media research on the other hand also has not paid much attention to the role of life events when studying the effects of social media use on well-being (see below for exceptions). The present paper is going to bring these two lines of research together.

Facebook use and adaptation to life events

To our knowledge, there are no studies using longitudinal panel data to examine the effect of Facebook use on the trajectories of adaptation to life events. There is work on internet use for coping with life events; Van Ingen, Utz, and Toepoel (2016) compared offline and online forms of coping and assessed also the use of social network sites as one form of online coping. They found that social network sites are used in several ways: for mental disengagement (e.g., distraction by looking at profiles of friends, watching funny content in the news feed), but also for problem-focused and socio-emotional coping. Interestingly, all three forms of online coping were negatively related to life satisfaction, although equivalent form of offline coping showed positive relationships with life satisfaction.

Bevan et al. (2015) examined how life events were shared on Facebook and found that negative events were more likely to be shared directly whereas positive events were often shared indirectly via photos. They did however not assess life satisfaction. The study that is most relevant for our paper, is the work by Burke and Kraut (2013) on Facebook use after job loss. They had three waves of panel data and could access log data to measure actual Facebook use. Burke and Kraut (2013) did not find changes in actual Facebook use, but directed communication with stronger ties led to an increase in social support, a decrease in stress in the long term and a higher chance of finding a new job. In contrast, passive reading of posts was associated with a decrease in social support. In a similar vein, Zhang (2017) found in a cross-sectional study that self-disclosure about stressful life events moderated the effect of stressful life events on satisfaction with life; people who disclosed more were less affected by stressors. Self-disclosure was also related to online social support; when including both predictors, the effect of self-disclosure was no longer significant and only online social support mattered. Thus, there are some first hints that Facebook use, especially active use, might be beneficial for the adaptation to life events.

Although there is not much work on Facebook use after life events, there are many studies on the effects of Facebook use on well-being in general (see Verduyn et al., 2017, for a review). Positive and negative effects have been found, and in a similar vein, positive and negative effects of Facebook use on adaptation to life events are possible. With regard to positive effects, Facebook users might adapt better to life events because they receive more social support from the social capital they maintain and build by using Facebook (Ellison et al., 2007, Ellison et al., 2014). There are two theoretical models how social support can affect adaptation to life events, which both have received empirical support: the buffer model and the direct effect model (Cohen & Wills, 1985; Holt-Lunstadt & Uchino, 2015).

According to the buffer model (Cohen & McKay, 1984), social support functions as a buffer; the drop in life satisfaction in the reaction phase during a stressful life event should thus be lower for people with a high level of social support. According to the direct effect model (Cohen and Wills, 1985, Pocnet et al., 2016), people with higher levels of social support experience also a drop in life satisfaction during the reaction phase, but recover more quickly and to a larger degree in the adaptation phase due to the social support they receive afterwards – either because they had higher levels of social support at the baseline or because they are more successful in leveraging their social capital, for example by asking their online friends for social support. Nabi, Prestin, and So (2013) found a stronger relationship between the number of Facebook friends and perceived social support for people who have experienced several life events supporting the idea that Facebook users mobilize their social support mainly in the adaptation phase.

Another line of theorizing assumes negative effects of Facebook use. Research has found that passively browsing the posts of Facebook friends can decrease well-being because the frequent upward social comparisons encountered when reading the mainly positive updates from friends trigger envy and depression (for reviews see Appel et al., 2016, Verduyn et al., 2017). During stressful life events such as financial hardship, seeing that others are able to go to fancy dinners or on vacation could result in more unfavorable upward social comparisons and thus even further decrease life-satisfaction.

The present research

The goal of the present research is to examine whether and to what extent Facebook use can help people to adapt to negative life-events, more specifically, the experience of financial hardship. Before turning to the exact processes, we first want to examine whether Facebook users and non-users differ in their reaction to the experience of financial hardship and the subsequent adaptation to the life event. As said above, we chose financial hardship as stressful life event because its prevalence in our sample was sufficiently high and because financial hardship is less closely related to age than for example widowhood or illness, decreasing the risk that age is related to both, the likelihood of experiencing the life event and social media use.

Although there is work on the relationship between Facebook use and well-being, there is hardly any work looking at the adaptation to life events and the role of Facebook use therein (see Burke & Kraut, 2013 for an exception). As outlined above, different lines of reasoning are possible. The social support models (Cohen & Wills, 1985) would argue that Facebook use helps people to adapt to life events, whereas the social comparison - envy work would predict negative effects (Appel et al., 2016). Therefore, we pose an open research question on the relationship between Facebook use and financial satisfaction after experiencing financial hardship:

: Do Facebook users and non-users differ in their adaptation to financial hardship?

In the next step, we zoom in into the role of online social support, first, because social support can have a direct as well as a buffering effect when people experience life events (Cohen & Wills, 1985) and second, because receiving social support from social capital built on social network sites is also the proposed mechanism why active Facebook use might contribute to higher well-being (Verduyn et al., 2017). In the present paper, we focus on online social support and not on offline social support because we want to examine the effects of Facebook use, which are most likely to happen online. Although online support can transfer into offline support in some cases (Trepte et al., 2015), an exclusive focus on offline support would disregard many forms of online social support, such as receiving useful information or emotional support (Van Ingen et al., 2016) and consequently lead to an underestimation of Facebook effects. Please note that our online social support measure covers all types of social support received online (regardless of platform) because an exclusive focus on online social support retrieved from Facebook would bias the results as well because non-Facebook users of course could not retrieve any social support from Facebook.

As described above, the buffer model (Cohen & McKay, 1984) and the direct effect model (Cohen and Wills, 1985, Pocnet et al., 2016) both expect positive effects of social support on adaptation to life events; the models differ mainly in the onset of this effect (already in the reaction phase or mainly in the adaptation phase). Although prior studies found weaker relationships between online social support and life satisfaction than between offline social support and life satisfaction (Liu and Yu, 2013, Trepte et al., 2015), Zhang (2017) found positive relationships between online social support on Facebook and life satisfaction after life events, indicating that online social support might have a stronger effect in stressful times. Zhang (2017) studied a student sample, so the life events were academic stressors, trouble with parents of friends or a change in living environment, but we expect to find also a positive relationship between online social support and financial satisfaction for people experiencing financial hardship.

Online social support is positively related to financial satisfaction in the adaptation phase.

It is unclear, whether potential effects of Facebook use on adaptation to life events are really due to a positive effect of Facebook use on social support or rather a selection effect such that people with higher levels of social support are more likely to use Facebook (a problem well known in studies of media effects, e.g. Jennings & Zeitner, 2003). If Facebook use is just a proxy for social support, the effects of Facebook use should no longer be significant once social support is included in the model. As prior research has not focused specifically on the role of Facebook use in the adaptation to life events, we pose an open research question:

Are the differences between Facebook users and non-users in adaptation to financial hardship still significant when controlling for online social support?

In a second set of analyses, we want to have a look only at the Facebook users and examine the role of passive and active Facebook use. As said above, prior research suggests that active use is beneficial for well-being because it increases social support (Verduyn et al., 2017, Zhang, 2017). Active use refers to posting status updates, commenting on the updates of others or exchanging private messages (Burke et al., 2011) – behaviors that are aimed at maintaining social relationships and suitable to elicit social support (Utz & Breuer, 2017; Verduyn et al., 2017). Burke and Kraut (2013) found positive effects of direct communication with strong ties after a job loss; we expect that these findings generalize also to the life event of financial hardship and active Facebook use in general.

Active Facebook use (posting) is positively related to financial satisfaction.

Passive use, defined as passively consuming social news by Burke et al. (2011), decreases well-being (Verduyn et al., 2017). Please note that passive use is not necessarily a low effort activity, although skimming of posts can create a feeling of ambient awareness, a sense of what is going on in one's network (Levordashka & Utz, 2016). Reading the positive updates in one's network can trigger social comparison processes which subsequently lead to envy and reduced satisfaction (Krasnova et al., 2013, Tandoc et al., 2015). We therefore expect the following relationship:

Passive Facebook use (reading) is negatively related to financial satisfaction.

With regard to posting, we expect to replicate the finding that posting is related to online social support (Utz & Breuer, 2017; Zhang, 2017).

Active Facebook use (posting) is positively related to social support online.

Zhang (2017) found a positive relationship between self-disclosure about negative life events on Facebook and perceived social support. Social support was also positively related to life satisfaction. However, the effect of self-disclosure on life satisfaction was no longer significant when controlling for social support. Thus, there is a first hint from cross-sectional data that social support might mediate the effect of Facebook use (Zhang, 2017). We explore whether we find a similar pattern in our longitudinal dataset.

Is the effect of posting on adaptation still significant when controlling for online social support?

Method

Sample and procedure

The data are part of a larger longitudinal study in which a representative sample of Dutch internet users was followed for four years [see XXX [blinded for review] for the complete list of variables per waves]. Participants received a survey every six month which assessed their social media use. The survey also contained a bunch of social capital indicators. In wave 1, N = 3367 people were surveyed; in wave 8, the sample consisted of 861 individuals who had participated in all seven earlier waves. Additionally, we had invited all participants from wave 1 who had dropped out in one of the subsequent waves to wave 8; 884 from those participated in the final wave.

For the present analyses, we used a subset of this sample. We selected people who experienced financial hardship in wave 2 or later (to have at least one wave as baseline) and recovered from this hardship in one of the subsequent waves. Thus, we distinguished between a baseline without financial difficulties, a reaction period in which financial difficulties occurred and an adaptation period in which the participants had to adapt to the life event. Participants who experienced no financial difficulties in one of the eight waves, participants who already experienced financial difficulties in wave 1 and did therefore not provide data for the baseline, or participants who did not provide data on the adaptation process were excluded. This procedure left us with n = 484 participants (2510 time points). From these, 228 were female and 256 male. Seventeen percent were between 18 and 29 years old, 15% between 30 and 39, 19% between 40 and 49, 27% between 50 and 64, and the remaining were older than 65. The mean net income was between 1900 and 2100 EUR. Roughly one third (35%) had lower education, 43% had medium education and 18% had higher education (based on the classification also used by Statistics Netherlands (CBS)). The average number of included data points per participant was M = 5.19 (Minimum = 3; Maximum = 8). Please note that (new) financial difficulties were allowed to occur within the adaptation period.

Measures

Life events. Participants indicated for a list of life events whether they had experienced them in the last six months. In the current paper, we focus on “Did you experience financial problems?” (0 = no; 1 = yes).

Financial satisfaction. Satisfaction with various aspects of life was measured with a shortened version of the life satisfaction scale by Priebe, Huxley, Knight, and Evans (1999). For the present paper, we focus specifically on financial satisfaction. This was measured with the item “How satisfied are you with your financial situation?”. We adapted the original answer categories (couldn't be worse, displeased, mostly dissatisfied, mixed, mostly satisfied, pleased, couldn't be better) to a 7-point Likert scale ranging from 1 = very unsatisfied to 7 = very satisfied.

Facebook use. Respondents were asked whether they use Facebook or a similar social network site (yes vs. no). Active and passive Facebook use was assessed by asking how often they posted or read/skimmed the messages of others on Facebook, respectively. Answers were given on a five-point scale (1 = “multiple times a day”, 2 = “once a day”, 3 = “a few times a week”, 4 = “a few times a month” 5 = “rarely”. These five categories were considered as sufficiently differentiated to capture active and passive Facebook use. Items were recoded so that higher values represent more frequent use.

Social support online. To measure online social support the UCLA (Dunkel-Schetter, Feinstein, & Call, 1986) was adapted so that it differentiated between online and offline social support (Utz & Breuer, 2017). Respondents indicated whether they received different types of social support from their partner, a close friend or family member (offline), from their partner a close friend or family member (online), acquaintances (offline), acquaintances (online), or people they only know online. The different types of online social support were collapsed across the three items into one scale measuring social support online; only these items were used in the present paper.

Control variables. We also assessed demographics (age, sex, education level, and income). Unfortunately, the marketing company that collected the data has changed the scale for assessing income during the eight waves from a quite fine-grained measure in steps of 200 EUR to a more global measure with only seven categories (e.g., modal income; slightly below modal income). We opted for the more fine-grained scale used in the first three waves to control for the effect of actual income on financial satisfaction.

See Table 1 for the descriptives at the baseline and the correlations.

Table 1

Descriptives and correlations at the baseline.

	M (SD)	1	2	3	4	5	6
1 Financial satisfaction	4.17 (1.30)	–
2 Facebook (0 = no, 1 = yes)	0.74 (0.44)	.05	–
3 Frequency of reading (n = 350)	3.71 (1.37)	.01	n.a.	–
4 Frequency of posting (n = 346)	2.23 (1.28)	-.05	n.a.	.40∗∗∗	–
5 Online social support	2.20 (0.89)	.12∗∗	.20∗∗∗	.28∗∗∗	.41∗∗∗	–
6 Sex (1 = female, 2 = male)	1.53 (0.5)	-.05	-.09∗	-.04	.20∗∗∗	.07
7 Income	9.48 (4.70)	.17	-.05	.05	.19∗∗∗	.15∗∗	.15∗∗

Note: ∗p < .05, ∗∗p < .01, ∗∗∗p < .001 Income at the baseline was measured on a scale from 1 = no income, 2 = 700 EUR, 2 = 700–900 EUR, …, until 20 = more than 4100 EUR.

Data analysis

For the analysis of the reaction and adaptation patterns of participants, Hierarchical Linear Models (HLMs) for longitudinal data (Raudenbush & Bryk, 2002) were estimated. On level 1, the HLMs included a model for individual change depending on time and time variant predictors; on level 2, the models included time-invariant predictors like sex or variables for which only one measurement is available. To analyze the adaptation process following the reaction to financial problems, we had to construct a time covariate displaying the growing temporal distance from the life event. In the baseline and the reaction period this variable was coded “0”, in the first period after the reaction it was coded “1”, in the second period after the reaction it was coded “2” and so on. We started with a baseline model, included the life event and the adaptation phase in model 2, checked the stability of the model when including demographics as control variables and added then Facebook use and social support. In the final step, we also examined the two-way interactions of Facebook use and online social support with life event/time from the event.

Within the subsample of Facebook users, after assessing the baseline models, we ran first a model with financial satisfaction as dependent variable in which we included frequency of active and passive use as well as their interactions with life-event and temporal distance from the life-event. Next, we used the same predictors to predict the level of online social support. In the final step, we included both, active and passive Facebook use and online social support, to predict financial satisfaction.

Results

Preliminary analyses

We first checked whether the finding that Facebook users show higher levels of online social support than non-users found for the full sample of Facebook users (Utz & Breuer, 2017), also held for the smaller subsample of the people who experienced financial hardship. Facebook users reported descriptively higher levels of online social support in seven out of the eight waves; and these differences were significantly higher in four of the eight waves (ts > 2.39, p < .05, respectively). We also checked for differences in income in the first three waves (before the measure was changed). Facebook users did not differ significantly from non-users and reported descriptively even slightly lower income levels, all ts < 1.48.

Hierarchical linear models

Facebook users vs. non-users. First, we ran a baseline model (see Table 2) without predictors in model 1 to examine how much variance in financial satisfaction is explained by differences between participants and within-person differences. The intraclass correlation of 0.52 indicated that 52% of the variance is explained by within-person differences, whereas 48% are due to between-person differences.

Table 2

Hierarchical linear models for the effects of Facebook use and social support on financial satisfaction.

Effects	Model 1	Model 2	Model 3	Model 4	Model 5
	Baseline model	Reaction & adaptation model	Including sex and income	Including Facebook use	Facebook + social support
Fixed Effects
Intercept	4.06∗∗∗	4.21∗∗∗	4.19∗∗∗	4.15∗∗∗	4.15∗∗∗
time-invariant predictors
Male (z-transformed)			−0.11∗∗∗	−0.10∗∗∗	−0.12∗∗∗
Netto income (mean over waves 1–3; z-transformed)			0.21∗∗∗	0.21∗∗∗	0.20∗∗∗
time-variant predictors/social media
Facebook use (Dummy)				0.06∗∗∗	0.06∗∗∗
Social support (z-transformed)					0.03∗∗∗
Reaction
Life event (Dummy)		−0.51∗∗∗	−0.52∗∗∗	−0.38∗∗∗	−0.36∗∗∗
IA life event X Facebook use				−0.22∗∗	−0.25∗
IA life event X social support					0.08∗∗
Adaptation
Temporal distance from life event		0.04∗∗∗	0.04∗∗∗	0.04 ∗	0.04∗
IA temporal distance X Facebook use				0.01∗∗∗	0.01
IA temporal distance X social support					0.03∗∗∗
Random effects
Residual	0.97∗∗∗	0.87∗∗∗	0.93∗∗∗	0.94∗∗∗	0.94∗∗∗
Variance of the intercept	0.91∗∗∗	0.93∗∗∗	0.84∗∗∗	0.81∗∗∗	0.80∗∗∗
Variance of temporal distance		0.01∗∗∗
Covariance between the intercept & temporal distance		−0.03∗∗∗
Model fit
−2 Restricted Log Likelihood	7895.94	7725.25	6447.67	6289.44	6287.00
Akaike's Information Criterion (AIC)	7899.94	7733.25	6451.67	6293.44	6291.00
Hurvich and Tsai's Criterion (AICC)	7899.94	7733.27	6451.68	6293.44	6291.01
Bozdogan's Criterion (CAIC)	7913.59	7760.56	6464.94	6306.65	6304.21
Schwarz's Bayesian Criterion (BIC)	7911.59	7756.56	6462.94	6304.65	6302.21

Note: IA = interaction; ∗p < .05, ∗∗p < .01, ∗∗∗p < .001.

In model 2, we added the life event and the temporal distance from the event. As can be seen at the bottom of Table 2, adding these predictors improved model fit – AIC, BIC and the other indicators of model fit were reduced. The negative estimate of the fixed effect of life event (−0.51) indicated that financial satisfaction dropped during the life event. The positive estimate for the fixed effect of temporal distance (0.04) indicated that people slowly adapted with increasing time from experiencing financial hardship (see Fig. 1).

Fig. 1

The effect of life event and temporal distance on the reaction and adaptation to financial difficulties.

In the next step, we explored the role of demographics (sex, age, education level) and income. Only the fixed effects of sex and income were significant; we continued in model 3 thus with the more parsimonious model containing only the significant controls. Not surprisingly, people with a higher income showed higher levels of financial satisfaction (0.21). Males had a slightly lower level of financial satisfaction than females (−0.11). The effects of life event in the reaction phase and temporal distance in the adaptation phase remained significant.

In model 4, we included Facebook use and the two-way interaction terms between Facebook use and life event and between Facebook use and temporal distance from the event as fixed effects.1 Facebook use had no effect on financial satisfaction. The estimate of the interaction between Facebook use and life event was however significant (−0.22). As can be seen in Fig. 2, Facebook users experienced a larger drop in financial satisfaction during the event, but showed higher levels of financial satisfaction in the adaptation phase. This pattern does not support the idea that Facebook acts as a buffer. In this case, the drop should have been smaller. The answer to RQ1 is therefore that Facebook users and non-users differ in their reaction to experiencing financial hardship; the Facebook users react more extremely on the life event, but adapt also quickly back to a higher set-point of financial satisfaction.

Fig. 2

The influence of Facebook use and life event on the reaction and adaptation to financial difficulties.

In the final model, we also added online social support and its interactions with life event and temporal distance. The effect of life event (−0.37) was still significant, but smaller than in the earlier models, whereas the effect of temporal distance remained the same (0.04), but was no longer significant. Online social support did not have a direct effect, but there was a significant interaction between online social support and temporal distance to the life event (0.03). As can be seen in Fig. 3 and in line with H1, the increase in financial satisfaction during the adaptation phase was steeper for people with high levels of online social support.

Fig. 3

The influence of social support (−/+ 1 SD) on the reaction and adaptation to financial difficulties.

Facebook use still interacted with life event, and the coefficient remained almost identical. The answer to RQ2 is thus that the effects of Facebook use and social support are additive; the interaction between Facebook use and life event is significant even when controlling for social support. Moreover, the effect of Facebook use is larger in the reaction phase, whereas social support matters more in the adaptation phase.

Subsample of Facebook users. To test hypotheses 2 and 3, we first ran two baseline models only with the subsample of Facebook users, one for financial satisfaction (Model 1) and one for social support (Model 2). Next, we ran a model that examined the effect of active (posting) and passive (reading) Facebook use on financial satisfaction. We also entered the interaction terms between posting respectively reading and life-event/distance from the event. As can be seen in Table 3, when comparing AIC, BIC and the other fit indices between Model 1 and Model 3, including the Facebook variables improved model fit. However, only for posting a significant interaction with distance from the life events emerged (0.04; Table 3, Model 3). As can be seen in Fig. 4, people who posted less frequently (- 1 SD) showed somewhat higher levels of financial satisfaction before and during the event. After recovering from the event, financial satisfaction stayed roughly on the same level. For people who posted frequently (+1 SD), financial satisfaction increased steadily with increasing distance from the life event. H2 is thus partly supported; active Facebook use is only positively related to financial satisfaction when the event happened a while back. In contrast to H3, there was no significant effect of passive use.

Table 3

Hierarchical models on the role of posting and reading (subsample of Facebook users).

	Model 1 Baseline model satisfaction	Model 2 Baseline model social support	Model 3 Facebook use => satisfaction	Model 4 Facebook use => social support	Model 5 Facebook use & social support => satisfaction
Fixed Effects
Intercept	4.05∗∗	2.26∗∗∗	4.20∗∗∗	2.30∗∗∗	4.20∗∗∗
time-invariant predictors
Male (z-transformed)			−0.12∗	0.07	−0.13∗
Netto income (mean over waves 1–3; z-transformed)			0.20∗∗∗	0.07	0.21∗∗∗
time-variant predictors/social media
Facebook use: posting (z-transformed)			−0.05	0.12∗∗∗	−0.01
Facebook use: reading (z-transformed)			−0.06	0.08∗∗	−0.05
Social support (z-transformed)					0.02
Reaction
Life event (Dummy)			−0.60∗∗∗	0.02	−0.61∗∗∗
IA life event X Facebook use: posting			0.04	−0.01	0.02
IA life event X Facebook use: reading			−0.05	−0.02	−0.05
IA life event X social support					0.07
Adaptation
Temporal distance from life event			0.04∗∗	−0.01	0.04∗∗
IA temporal distance X Facebook use: posting			0.04∗	0.02	0.02
IA temporal distance X Facebook use: reading			0.03	−0.02	0.03
IA temporal distance X social support					0.03†
Random effects
Residual	1.05∗∗∗	0.31∗∗∗	0.97∗∗∗	0.29∗∗∗	0.95∗∗∗
Variance of the intercept	0.90∗∗∗	0.48∗∗∗	0.82∗∗∗	0.41∗∗∗	0.83∗∗∗
Model fit
−2 Restricted Log Likelihood	5646.74	3713.68	4421.74	3006.24	4585.15
Akaike's Information Criterion (AIC)	5650.74	3717.68	4425.74	3010.24	4589.15
Hurvich and Tsai's Criterion (AICC)	5650.74	3717.68	4425.75	3010.24	4589.16
Bozdogan's Criterion (CAIC)	5663.67	3730.61	4438.20	3022.78	4601.68
Schwarz's Bayesian Criterion (BIC)	5661.67	3728.61	4436.20	3020.78	4599.68

Note: IA = interaction; ∗p < .05, ∗∗p < .01, ∗∗∗p < .001.

Fig. 4

The influence of posting (−/+ 1 SD) on the reaction and adaptation to financial difficulties.

To test H4, we examined in a similar vein whether reading and posting were related to social support (Table 3, Model 4). Again, when comparing with the baseline model (Model 2), model fit improved. In line with H4, the main effect of posting was significant: People who posted more frequently also received more social support online (0.12). Interestingly, a similar, but smaller positive (0.08) effect occurred for reading.

In a final step and to answer RQ3, we added online social support in the model with financial satisfaction as predictor (Model 5 in Table 3). The interaction term between temporal distance from the life event and Facebook use was no longer significant. However, social support was also not related to financial satisfaction; the interaction with temporal distance was only marginally significant. The answer to RQ3 is that the effect of Facebook use is no longer significant once social support is added. However, model fit was worse when social support was included compared to Model 3 with only posting and reading were used as predictors; these results should therefore be treated with caution.

Discussion

The goal of this study was to examine whether Facebook use can help people to adapt to negative life events, more specifically, the experience of financial hardship. The analysis revealed that Facebook users showed more extreme reactions to the life event than non-users. After the life event, they showed higher levels of financial satisfaction than non-users did, but during the life event, they showed lower levels of financial satisfaction than non-users. There was also an effect of social support: People who reported higher online social support showed a steeper increase of financial satisfaction in the adaptation phase. Within the subsample of Facebook users, we found that posting and reading were both positively related to online social support. However, only posting was related to financial satisfaction, and mainly so in the later adaptation phase. This interaction was no longer significant when controlling for online social support, but social support had only a marginal effect on financial satisfaction in the adaptation phase and model fit decreased, indicating only weak support for a mediation effect.

These results have several theoretical and practical implications. The more extreme reactions of Facebook users show that Facebook use is not a buffer that prevents people from a drop in life satisfaction during a life event. The relatively larger drop could be caused by upward social comparisons. Many people share pictures of positive life experiences such as dinner with friends on Facebook (Krasnova et al., 2013). Research has shown that low perceived control (i.e., difficulties to get the envied object) is a precondition of envy (Smith, 2004). Suddenly realizing that one can't afford fancy dinners and weekend trips anymore might trigger higher levels of envy that explain the larger drop in financial satisfaction. We did not find a significant effect of frequency of reading on financial satisfaction, but this does not necessarily indicate that the social comparison explanation does not hold. We only assessed frequency of reading, but perceived upward social comparisons while reading might have been a better predictor that should be used in future studies.

The results are also informative for work on (online) social support as they show that social support helps even more in the adaptation phase than in the reaction phase. Again, the pattern of the interaction speaks against the buffer hypothesis and favors the direct effect model: Both groups experienced a similar absolute drop in financial satisfaction as indicated by the almost parallel lines of financial satisfaction in Fig. 3. This finding also has a methodological implication as it points to the importance of longitudinal studies including a baseline condition – cross-sectional studies that only look at the reaction phase might easily come to the conclusion that social support is a buffer because people with higher levels of social support report higher financial satisfaction during the reaction phase than people with lower levels of social support (see Anusic & Lucas, 2014, for a similar argument).

Our results indicate that the effects of online social support become more pronounced over time: Whereas people with low levels of social support adapted to the baseline level of financial satisfaction, people with high levels of social support showed a further increase in financial satisfaction. Burke and Kraut (2013) reported a similar pattern in their study on Facebook use after job loss; their qualitative data showed that people often received supporting comments when they already had found a new job. Prior inconsistent effects have been explained with the fact that people often receive more social support in stressful times, but that this higher level of social support is often still less than they expected or not exactly what they expected (“support deterioration”, Kaniasty, 2012). This might explain the relatively small effects of online social support we found during and shortly after the life event.

Within the group of Facebook users, frequency of posting was positively related to online social support. This is in line with prior research (Verduyn et al., 2017, Zhang, 2017). Surprisingly, reading was also positively related to online social support. This is in contrast to Burke and Kraut (2013) and surprising because reading is usually considered as detrimental to well-being due to the exposure to upward social comparisons. A possible explanation could be that receiving social support requires both, telling one's friends that one needs social support and actually reading what friends contribute. The mixed consequences of reading – unfavorable social comparisons and reading social support messages – might also explain why we did not find a significant relationship of reading with financial satisfaction.

Utz & Breuer (2017) had found no direct effect of Facebook use on well-being when analyzing the first six waves of the longitudinal study. The rationale for looking at the moderating role of life events was to examine whether Facebook use especially helps in times of negative life events. The current results indicate rather the opposite: Active Facebook use contributes to well-being, but the more the longer the event lies back. An important theoretical contribution is thus that we identified a moderator of the effects of Facebook use on well-being: life events. It seems that Facebook use can further increase happiness in good times, but does not result in higher life satisfaction in times of crisis. Whereas a shared joy seems to be double joy, a shared problem seems not to be a problem halved. Future research should examine whether this pattern holds also for other life events.

Our study makes also an important contribution to research on adaptation theory. To our knowledge, it is the first that looked at the role of media use. We focused only on one specific life event and looked also only at Facebook use, but the results show that media use also affects the set-points of people and the impact of life events. Future studies should therefore pay attention to media use.

A limitation of our study is that we focused only on financial hardship and can thus not say whether the results generalize to other types of life events such as a relationship break up, a severe illness or widowhood. For financial hardship, tangible support such as lending money could be more important than for example comfort or esteem support. We selected a life event that is unequivocally negative (a relationship break-up can also be positive for some people) and that occurred relatively frequently in our sample. Power constraints withheld us therefore from looking at the impact of other life events. The role of Facebook in recovering from a break-up might be different and/or moderated by Facebook use of the ex-partner. Studies have shown that being confronted with an ex on Facebook can hamper adjustment (Tran & Joormann, 2015); it might thus take longer until Facebook users report higher levels of adaptation than non-users. Another limitation is that we measured financial satisfaction only with one item. A strength is that we used several waves of a longitudinal panel study to examine the reaction and adaptation phase. Moreover, we disentangled Facebook use and online social support. The results showed interesting temporal dynamics that can be the starting point for a promising line of research.