Abstract
Recent literature has suggested that smartphone addiction is negatively associated with users’ psychosocial well-being. Much of the research on this subject, however, is of a correlational nature, which has been pointed out as an important limitation that does not allow distinguishing the antecedents of the consequences. In this study, 416 smartphone users were followed for 1 year (three waves separated by 6 months each) to assess the relationship between smartphone addiction and social support. Cross-lagged model results indicated that social support predicts later addiction to the smartphone and that smartphone addiction decreases social support over time. Growth mixture model results indicated that the decrease in social support during the follow-up year was higher for users with greater smartphone addiction at the beginning of the study. Multivariate and univariate analyses of variance indicated that some personal characteristics of users (extroversion, neuroticism, and sensation-seeking) could affect the evolution of social support related to smartphone addiction. In general, these results suggest that the extensive use of a social communication technology such as the smartphone could have the paradoxical effect of diminishing the psychosocial well-being of its users.
Keywords
In the present study, we analyze the relationship between the excessive use of the smartphone, a social technology that allows social connectivity in real time, and the potential negative implications it has on the social world of the user. To that end, we followed-up during a year to 416 smartphone users and we studied the relationship between their smartphone addiction and social support over time as well as its association with the psychological distress of users.
Back in 1998 Kraut, Patterson, Lundmark, Kiesler, Mukophadhyay, and Scherlis, launched the idea that social technologies, specifically designed to enhance real-time connectivity, had the paradoxical effect of causing isolation and deteriorating the psychological well-being of their users. Although a few years later, they published new studies that seemed to contradict the first results, the influential idea of a potential paradox of social technologies attracted the attention of researchers. In their original work, Kraut et al. (1998) studied Internet and the use of computers to connect to the Internet. Nowadays, studies on communication technologies and the well-being of users have been moving toward the study of smartphones for a number of reasons. A smartphone is basically a computer that allows audio and video calls but for its small size, its versatility, as well as its connectivity, among others, the smartphone has extended its use among the population (Lee, 2014). In Spain, where the present study was undertaken, smartphone ownership in the adult population is among the highest in the world and the first one in the European Union (Deloitte, 2015; Poushter, 2016). Due to the increasing number of applications (communication, health, entertainment, shopping, financial, weather forecasting, etc.), it is foreseeable that its use will continue to grow in the population (Poushter, 2016). But this extensive use hides some issues that have not gone unnoticed to experts. Of particular relevance from the point of view of the psychological well-being of the users is the extensive use and the potential addiction to the smartphone. As Billieux, Maurage, Lopez-Fernandez, Kuss, and Griffiths (2015a) have recently discussed, several terms are frequently used to describe the potentially harmful use of the smartphone, including among others mobile phone (or smartphone) addiction and problematic mobile phone use (PMPU). Although these two terms are sometimes used interchangeably, Billieux et al. (2015a) have argued that PMPU is not always associated with an addiction. For example, PMPU may lead to antisocial (e.g., aggressive phone calls) or risky behaviors (writing messages while driving) that do not necessarily fit into the description of addictive behavior. In the present study, we specifically focus on the problematic use of smartphones that may be regarded as an addiction.
Research about the antecedents and potential consequences of smartphone addiction (Kim, 2017) has been nourished by the major findings on addiction to the Internet and to mobile phones. This type of behavioral addiction is known to be associated with psychosocial outcomes such as social isolation, psychological adjustment problems, conflict in family relationships, mental health risk (Elhai, Dvorak, Levine, & Hall, 2017), or higher rates of malware and risky attitudes toward smartphone use (Herrero, Urueña, Torres, & Hidalgo, 2017b). According to Billieux et al. (2015b), this type of behavioral addiction is closely related to the necessity to maintain relationships and obtain reassurance from others. In this vein, there is an increasing number of research analyzing the link between the addiction to the smartphone and the social relationships of users.
Social Support, Smartphone Addiction, and Psychological Well-Being
The literature on the relationship between addiction to smartphones and social support is scarce. There is some evidence, however, about the negative role that the social relations of the user might have on the addiction to smartphone. Thus, in a pioneering study conducted by Tossell, Kortum, Shepard, Rahmati, and Zhong (2015), they found that addiction to smartphone was positively related to a greater use of social applications like Facebook, or instant messaging, further suggesting that smartphone addiction is related to the social environment of the user. Thirty-four users were provided with smartphones that logged all their use over the course of a year and found that smartphone usage was more fragmented in the case of addicted users, thus suggesting a need for constant access to information about user’s social relationships that, in the end, may become difficult to control over time. This type of behavior in smartphone-addicted users would be consonant with their need to obtain reassurance in affective and close relationships (Billieux et al., 2015b). In this sense, users would develop addictive behaviors because of their constant need to obtain approval and support of their social relationships (van Deursen, Bolle, Hegner, & Kommers, 2015). According to this, lower levels of social support would be related with an increase in these types of behaviors, potentially leading to addiction to the smartphone (Lu et al., 2011). This is consistent with empirical evidence showing that difficulties in social relationships might be responsible for smartphone addiction (Kim, 2017; Kwon, So, Han, & Oh, 2016). Most of these studies, however, do not allow to discard alternative explanations. Given the correlational nature of these studies, it could also be argued that levels of social support may be the result of excessive use of smartphones, resulting in social isolation and low perceived social support (World Health Organization [WHO], 2015). In spite of these limitations, the literature has not paid attention to the potential effects that the addiction to the smartphone could have in the social support of the users. Rather, it has traditionally been understood that (low) social support could be a precursor to smartphone addiction, as noted above. To our knowledge, the study of the opposite relationship—smartphone addiction influencing user’s social support—has been largely neglected in the literature.
There is some indirect evidence to suggest that, in fact, social support might be negatively affected by smartphone addiction. For example, Lemmens, Valkenburg, and Peter (2011) found that behavioral addictions (i.e., gambling) increase the feeling of loneliness over time and that mobile phone can have detrimental effects on various spheres of daily living (Billieux, 2012), including an increase in conflicts with family and friends (De-Sola, Rodríguez, & Rubio, 2015), thus potentially leading to a decrease in social support. From this point of view, the social support of users addicted to the smartphone could be negatively affected.
The fact that smartphone addiction could have a negative effect on social support levels would have a significant detrimental influence on the user’s psychological health. The literature on the positive effects of social support on psychological well-being has been widely known for more than three decades (Cohen & Syme, 1985), so users with higher levels of addiction may be at greater risk of their psychological well-being (Choi, Lee, & Ha, 2012; Dan, Bae, Koo, Wu, & Kim, 2015; Davey & Davey, 2014; WHO, 2015). This line of thought was exemplified in Kraut et al.’s (1998) pioneering work on the extensive use of the Internet and the social support of users where greater use of the Internet was associated with declines in the size of their social circle as well as an increase in their depression and loneliness. Applied to the study of smartphones, it could be thought that real-time connectivity of smartphones could have a similar effect on levels of social support of users: a decrease in levels of social support and an increase in psychological distress.
In line with the scientific literature reviewed above, the general objective that guides the present research is the study of the relationships between smartphone addiction and social support over time and its potential influence on the psychological well-being of users. In accordance with the above-mentioned literature, we hypothesized that:
While Hypotheses 1 and 2 are closely related, they refer to different phenomena. The Hypothesis 1 tries to deepen our knowledge about the relationship between the levels of addiction to the smartphone and the levels of social support. It serves, therefore, to contrast the empirical evidence of correlational studies that indicate that both variables are negatively related. The time dimension is included here to account for their reciprocal influences over time. The Hypothesis 2 analyzes the possible trajectories of social support over time in their relationship with previous levels of addiction to the smartphone. It analyzes, therefore, to what extent previous levels of addiction to the smartphone influence not only the evolution of social support over time (intraindividual variability) but also seeks to analyze the existence of different trajectories in groups of individuals (interindividual variability).
To empirically test for these theoretical assumptions, we used data from 416 smartphone users whose social support were followed-up for a year and assessed in three different occasions (waves) every 6 months. Consistent with previous research regarding smartphone addiction and personality and psychological characteristics of the users (Aljomaa, Qudah, Albursan, Bakhiet, & Abduljabbar, 2016; Andreassen et al., 2013; Bianchi & Phillips, 2005; Billieux et al., 2015b; Butt & Phillips, 2008; Chen et al., 2016; Fullwood, Quinn, Kaye, & Redding, 2017; Hawi & Samaha, 2017; Kruger & Djerf, 2017; Lepp, Li, Barkley, & Salehi-Esfahani, 2015; Lian, You, Huang, & Yang, 2016; Lu et al., 2011; Randler, Horzum, & Vollmer, 2014; Stieger, Burger, Bohn, & Voracek, 2013; Wang, Ho, Chan, & Tse, 2015; Yao, He, Ko, & Pang, 2014; Zhitomirsky-Geffet & Blau, 2016), we also included an evaluation of the personality characteristics of individuals. These investigations have shown that a significant part of the addictive behavior could be directly explained by the user personality characteristics. From this point of view, to study the relationship between smartphone addiction and social support, it is therefore necessary to take into account the potential effect of user’s personality on addiction. To see to what extent the potential decline in social support over time linked to smartphone addiction had, in fact, effects on the psychological well-being of users, we also evaluated the psychological distress of users.
We also included sociodemographic variables in our study since there are previous research that has shown that levels of smartphone addiction and social support vary across age, gender, educational attainment, and size of locality (de-Sola, Talledo, de Fonseca, & Rubio, 2017; Gracia & Herrero, 2004; Herrero, Meneses, Valiente, & Rodríguez, 2004; Lopez-Fernandez et al., 2017). Finally, to better control for potential response bias, social desirability was also taken into account.
Method
Participants
Three waves of data from the Cybersecurity and Confidence in Spanish Households national survey conducted by the National Observatory of Telecommunications and Information Society of the Spanish Ministry of Industry were used for this study (see Herrero, Urueña, Torres, & Hidalgo, 2017a, 2017b, for a detailed description). We used self-reported data from 416 Internet users who were assessed in three occasions, separated from each other by an interval of 6 months, during the period July 2015 to December 2016.
Participants belonged to a representative sample of the Spanish population of Internet users 15 years old and over with residential Internet access. Primary sampling units were households and secondary sampling units were individuals within households. First, a Spanish representative sample of households in terms of autonomous communities, size of locality, social class, and number of persons in household were selected. Second, Internet users 15 years old and over within households were identified and selected. Initially, 661 users completed the questionnaire at Wave 1. From these 661 users, a total of 416 users (63%) provided complete data at both Wave 2 (after 6 months) and Wave 3 (a year later).
Variables and Scales
Sociodemographic variables
Sociodemographic variables were sex (male = 52.2%, female = 47.8%), age in five age-groups years (15–24 years [5.1%], 25–34 years [24.7%], 35–44 years [39%], 45–54 years [23.5%], and more than 55 years [7.7%; M = 3.27, SD = 1.07]), educational background (highest educational level attainment, 1 = elementary [1.5%], 2 = secondary [43.7%], and 3 = university studies [54.8%; M = 2.23, SD = 0.51]), and size of locality (from 1—less than 10,000—to 6 more than 500,000 inhabitants—M = 3.86, SD = 1.79). The distribution of localities was as follows: 1 = under 5,000 inhabitants (9.9%), 2 = 10,000–50,000 inhabitants (20.3%), 3 = 50,000–100,000 inhabitants (9.9%), 4 = 100,000–250,000 inhabitants (11.7%), 5 = 250,000–500,000 inhabitants (23.3%), and 6 = more than 500,000 inhabitants (24.9%).
Social support
The strong-tie support scale (Lin, Dean, & Ensel, 1986) was used to measure social support from intimate and confidant relationships with 3 items in a 4-5 point scale from 1 = never to 5 = most of the time. It measures to what degree respondents felt their support needs from close companions were fulfilled. The 3 items of the strong-tie support scale is a recommended measure of social support for large-scale surveys due to its brevity (Gracia, Herrero, Lila, & Fuente, 2009). Social support was measured in all three waves (see Table 1). Cronbach’s α was adequate for the three waves of data (α’s ≥ .65) in line with other research (Herrero & Gracia, 2011).
Descriptive Statistics of Variables of the Study in All Three Waves.
Note. N = 416.
To further analyze the potential correlates of the variation in social support over time, smartphone addiction, personality, and risk propensity were evaluated in Wave 1. Smartphone addiction was also evaluated in Wave 3 at the end of the study along with a measure of psychological distress (depressive mood).
Extent of smartphone addiction
We applied Bian and Leung’s (2015) criterion to evaluate smartphone addiction, using information from 8 items of their Smartphone Addiction Symptoms Scale that were most conceptually equivalent to Young’s screening instrument in Internet addiction (see Table 1; see Bian & Leung, 2015, for a more detailed explanation). Although those items were originally coded into five category responses from 1 = never to 5 = most of the time, to evaluate addiction, Bian and Leung considered only category responses 4 (many times) and 5 (most of the time) of these 8 items. To this end, all 8 items were first dichotomized (1–3 recoded to “0”, and 4–5 recoded to “1”) and summed, which yielded a value ranging from “0” to “8.”
Personality
An abbreviated version of the Big Five Inventory (BFI; Rammstedt & John, 2007) was used to measure personality traits of the Big Five model. The abbreviated version of the BFI uses 10 items with category responses ranging from 1 = completely disagree to 5 = completely agree. Personality was measured in Wave 1 (see Table 1)—Cronbach’s α ≥ .65. Authors found that this brief inventory retained significant levels of reliability and validity when compared to other measures of the Big Five personality factors such as the NEO Personality Inventory. The Brief Sensation Seeking Scale (Hoyle, Stephenson, Palmgreen, Lorch, & Donohew, 2002) is a 4-item scale that measures trait sensation seeking. Items responses ranged from 1 = strongly disagree to 5 = strongly agree (M = 2.37, SD = 0.77; Cronbach’s α = .82). Sensation seeking was measured in Wave 1.
Depressive mood
We used a 7-item validated version of the original 20 items Center for Epidemiological Studies Depression (CESD; Herrero & Gracia, 2011). The CESD-7 Scale is a short self-report scale designed to measure depressive symptomatology in the general population. The items of the scale are symptoms associated with depression. The CESD-7 was measured in Wave 3 (Cronbach’s α = .84).
Social desirability
The Strahan and Gerbasi (1972) short form of the Marlowe–Crowne Social Desirability Scale (Crowne & Marlow, 1960) was used. This scale consists of 10 true–false items of the original 33 items scale (1 = true, 2 = false). Negative items were reversed, so that higher scale scores reflect greater levels of social desirability. Social desirability was measured in Wave 1 (see Table 1).
Attrition Analyses
Close inspection of dropouts across sociodemographic variables revealed that dropouts (M = 2.74, SD = 1.03) were younger than nondropouts, M = 3.04, SD = 1.00; F(1, 659) = 10.14, p < .001. No significant differences were found for sex, χ2(1) = 0.62, ns, Cramer’s V = .03, ns; size of locality, F(1, 659) = 3.10, ns; and educational attainment, F(1, 659) = 0.56, ns.
As for the outcome variables and covariates of the study, we found that dropouts (M = 3.41, SD = 0.91) significantly scored lower on social support at Wave 1 than nondropouts, M = 3.59, SD = 0.85; F(1, 659) = 6.05, p = .014. Also, addiction to smartphone at Wave 1 was higher for dropouts (M = 1.14, SD = 2.07) than for nondropouts, M = 0.96, SD = 1.56; F(1, 659) = 10.01, p = .002. These data also suggested that the variability of social support and smartphone addiction was greater among dropouts than nondropouts. We estimated the Pearson’s bivariate correlation between these two variables for each of the groups. The reduction of the variability in the nondropouts group resulted in a lower covariation between social support and smartphone addiction in Wave 1, as seen in the Pearson’s correlations for dropouts (r = −.32) and nondropouts (r = −.20; ps < .001).
With regard to covariates of the study, personality measures differed across groups, Wilk’s λ (6, 654) = .97, p = .003, specifically in neuroticism, F(1, 659) = 17.09, p < .001: dropouts (M = 2.90, SD = 0.82), nondropouts (M = 2.59, SD = 0.96). Finally, no differences across groups of dropouts and nondropouts were found for social desirability at Wave 1, F(1, 659) = 0.16, ns.
Overall, younger participants and those scoring higher on neuroticism showed a tendency to abandon the study. Those participants in Wave 1 with lower levels of social support and those with higher levels of smartphone addiction also showed a tendency to abandon the study.
Analytical Strategy
In this study, we analyzed the relationship between smartphone addiction and social support in a number of complementary ways. First, we explored whether previous levels of smartphone addiction and social support at Wave 1 would predict levels of smartphone addiction and social support at Wave 3. To that extent, we estimated a cross-lagged panel model using EQS 6.3 software (Bentler, 2006). A cross-lagged panel model is a type of structural equation model that is used where two or more variables are measured at two or more occasions and interest is centered on the associations with each other over time. In our study, social support and smartphone addiction at Wave 1 were predictive of both social support and smartphone addiction at Wave 3 (see Figure 1). While cross-lagged results should not be interpreted in terms of causality, they may suggest the predominant direction of effects over time (Nylund, Asparouhov, & Muthén, 2007).
Cross-lagged model of the relationship between smartphone addiction and social support over time.
As shown in Figure 1, social support in Wave 1 predicts smartphone addiction in Wave 3, beyond the fact that there is a relationship between social support and smartphone addiction both in Waves 1 and 3. In addition, the effect of social support on smartphone addiction over time already takes into account the relationship between smartphone addiction in both Waves 1 and 3. The same applies to the interpretation of the relationship between smartphone addiction in Wave 1 and social support in Wave 3.
For the analysis of the evolution of social support and its relationship with the smartphone addiction we used Mplus 7 (Muthén & Muthén, 2012) to estimate latent growth mixture models. Although latent growth models are an interesting approach to the study of the evolution of a variable over time, this type of analysis assumes that a single growth trajectory can adequately approximate an entire population. Our interest, however, was the analysis of the effect of smartphone addiction in the evolution of social support, so an approach based on mixture models was preferred. Specifically, we estimated different models to study the relationship between levels of addiction to the smartphone and the evolution of social support. First, we estimated a conditional model with a single class or trajectory that maintained that all social support of the subjects varied uniformly over time in their relationship to smartphone addiction. Conditional growth mixture models that include a covariate has been shown to provide more information to refine the membership classification and may produce a more reliable solution (Muthén, 2004) if the covariate significantly influence the growth trajectory.
However, as Huang, Brecht, Hara, and Hser (2010) have noted, the inclusion of a covariate in the model may also increase the complexity of model specification and difficulty in parameter estimation. Because of this, the selection of appropriate covariates is key and must be done on theoretical grounds. In our study, levels of smartphone addiction in Wave 1 were entered in the model as a covariate to test for Hypothesis 2: The trajectories of social support over time among individuals are negatively influenced by previous smartphone addiction levels.
To do this, smartphone addiction at Time 1 predicted levels of the latent construct. Next, alternative models that included a growing number of classes or social support trajectories in their relationship with smartphone addiction were estimated. In these cases, with more than one class or trajectory, the coefficient of smartphone addiction to the latent class represents the multinomial logistic regression of smartphone addiction on the latent classes or trajectories. The model that best represented the data was selected through the comparison of model fit information, entropy, and likelihood ratio tests for competing models. Following Nylund, Asparouhov, and Muthén’s (2007) previous work on selecting competing growth mixture models, we relied on the Akaike information criteria and Bayesian information criterion and the bootstrap likelihood ratio for model selection. Because deciding on the number of classes may be difficult, however, we also considered the research questions, the substantive meaning of each solution, parsimony, and/or theory (Bauer & Curran, 2003). Additionally, we analyzed how clearly distinguishable the classes were based on how distinctly each individual’s estimated class probability was (as measured by entropy, ranging from 0 to 1 and with values close to 1 indicating clear classification) to test for the plausibility of each solution. After finding the model that best fit the data, we analyzed the relationships between social support trajectories and the remaining variables of the study. In this phase, we used multivariate analysis of variance (MANOVA) and univariate ANOVA using the statistical package SPSS 24 from IBM.
Results
Social Support and Smartphone Addiction Over Time
Cross-lagged effects
For the estimation of cross-lagged effects between smartphone addiction and social support across time, we tested a saturated model (with no degrees of freedom). In this model, both exogenous variables (those measured at Wave 1) and endogenous variables (all measured at Wave 3 and explained by the exogenous variables) are correlated. Results of this model (see Figure 2) indicated that, as expected, smartphone addiction and social support correlated negatively in both waves: Wave 1, r = −.19, p < .001; Wave 3, r = −.12, p < .01. Also, both measures were positively related over time: smartphone addiction in Wave 1 and Wave 3 (β = .59, p < .001); social support in Wave 1 and Wave 3 (β = .42, p < .001). The substantive part of the model indicated that social support in Wave 1 was negatively related to smartphone addiction in Wave 3 (β = −.10, p < .05) and that smartphone addiction in Wave 1 was negatively related to social support in Wave 3 (β = −.09 p < .05), beyond the influences of previous levels of social support and smartphone addiction in each case. And, also, controlling for the fact that levels of social support and smartphone addiction in Wave 3 are negatively correlated. We further estimated this cross-lagged model for men and women to test for the invariance of these results across sex. In this model, we constrained all paths and correlations to be equivalent across groups (for men and women). Model fit was adequate: Satorra-Bentler (S-B) χ2 = 3.40, df = 6, p = .75, thus suggesting that the effects were equivalent for men and women. No evidences of improving model fit through a significant reduction in χ2 were found.
Standardized results of cross-lagged model among the relationship between smartphone addiction and social support over time. *p < .05. **p < .01. ***p < .001.
Testing the Plausibility of Different Trajectories
While these results strongly suggested that levels of smartphone addiction and social support were also negatively related across time, they failed to inform about the potential influence of smartphone addiction on the evolution of social support over time. We used latent growth mixture modeling to further analyze these potential relationships. Several competing models were tested, from one-class to four-class solutions (see Table 2). Although the one-class solution outperformed in terms of BIC, the two-class solution showed better fit according to AIC and a similar fit with regard to the sample-adjusted BIC (see Table 2). In this model, however, the bootstrap likelihood ratio test was significant (p < .05), suggesting that the two-class solution was better that the one-class solution. The distance between classes, additionally, was high (entropy > .90), further suggesting that the two-class solution was tenable. Given the above, the model with two classes or trajectories was preferred over a model with a single class or trajectory. A solution of three and four classes or trajectories were also estimated (see Table 2), but they were considered inappropriate since they maintained classes with very low number of participants (a class with n = 11 for the three-class solution and two classes with less than 13 participants in the four-class solution). Class 1 grouped 40.6% of participants. Most likely latent class membership was of 94%. For this group, social support at Wave 1 was high (M = 4.42, p < .001) with an overall decrease in social support over time (growth slope = −.30, p = .02). Class 2 (59.4% of participants) was around the sample mean in social support at Wave 1 (M = 3.59, p < .001) and showed no change in social support over time (growth slope = 0.03, ns). Most likely latent class membership for this group was 96%. Smartphone addiction in Wave 1 had a significant effect both on social support at Time 1 (b = −.29, p < .001) and change in social support over time (b = .09, p = .02). Thus, smartphone addiction was related to lower levels of social support at Wave 1 and a steeper growth slope (b = .07, p = .02). In other words, participants scoring higher on smartphone addiction at Wave 1 decreased at a higher rate their levels of social support over time.
Growth Mixture Modeling Analysis: Results of Competing Models.
Note. N = 416. AIC = Akaike information criterion; BIC = Bayesian information criterion; BLRT = parametric bootstrapped likelihood ratio test.
Trajectories of Social Support
Latent growth mixture classification results obtained in Mplus were further submitted to univariate ANOVA and MANOVA using the statistical package SPSS 24. Results showed that gender, F(1, 411) = 1.14, ns; age, F(1, 411) = 0.79, ns; size of locality, F(1, 411) = 0.94, ns; and educational level, F(1, 411) = 2.51, ns, were equivalent across classes.
Also, this classification of participants seemed not to be affected by social desirability at Wave 1 and Wave 3 according to results of the MANOVA: Wilk’s λ = 0.99, F(2, 410) = 1.31, ns. As for the characteristics of personality, the MANOVA’s results showed significant differences between the two groups: Wilk’s λ = 0.94, F(6, 406) = 4.09, p < .001. Univariate analysis (ANOVA’s) showed that participants in Class 1 scored higher in extroversion, M1 = 3.28, M2 = 3.08, F(1, 412) = 3.79, p < .05; neuroticism, M1 = 2.73, M2 = 2.51, F(1, 412) = 5.66, p < .05; and sensation seeking, M1 = 3.42, M2 = 3.15, F(1, 412) = 12.97, p < .001, than participants in Class 2. These results suggested that personality had a significant effect on trajectories of social support over time in relation to smartphone addiction.
Also, those participants experiencing a decrease in social support due probably to their smartphone addiction (Class 1) showed higher levels of psychological distress in Wave 3 (depressive mood) than participants having a profile of low smartphone addiction who did not experience a decrease in social support (Class 2): M1 = 1.78, M2 = 1.62, F(1, 412) = 7.50, p < .01.
With regard to levels of smartphone addiction in Wave 3, ANOVA’s results indicated that there was also a significant difference in levels of smartphone addiction in Wave 3 between classes, M1 = 1.67, M2 = 0.38, F(1, 412) = 166.69, p < .001.
Discussion, Strengths, and Limitations
Discussion
Despite the growing interest in knowing more about the risk factors and consequences associated with addiction to smartphone, longitudinal studies on the subject are practically nonexistent. In this article, we present the results of a 12-month follow-up (in three waves) of 416 smartphone users in order to study the relationship between their psychosocial adjustment and their smartphone addiction. To achieve this goal, several analyses were carried out including cross-lagged causal models, latent growth mixture models, and both MANOVA and univariate ANOVA. Overall, results of the study provided empirical support for the three hypotheses of the study: smartphone addiction and social support are negatively related over time (Hypothesis 1), the trajectories of social support over time among individuals are negatively influenced by previous smartphone addiction levels (Hypothesis 2), and that this decrease in social support is in turn related to higher levels of psychological distress (Hypothesis 3).
With regard to the first of the hypotheses, the results of the cross-lagged causal models showed that higher levels of social support were predictive of lower levels of smartphone addiction in Wave 3, after accounting for previous levels of smartphone addiction in Wave 1. The same applied for the prediction of social support over time: Higher levels of smartphone addiction in Wave 1 were predictive of lower levels of social support in Wave 3. Recent literature on this topic has suggested that low levels of social support are likely to lead users to extensive (and probably addictive) use of the smartphone in order to obtain reassurance in affective and close relationships (Billieux et al., 2015b; Herrero et al., 2017a; van Deursen et al., 2015; Wang et al., 2015). Our results would corroborate this hypothesis, with the addition that the influence of social support on smartphone addiction has been analyzed in our study using temporal panel data. Related to this, our study also allowed us to analyze the consequences of smartphone addiction in social support. To the best of our knowledge, this is the first study that links previous levels of addiction to the smartphone with levels of social support over time.
While other correlational studies had suggested that social support might be negatively affected by smartphone addiction (Bian & Leung, 2015; Billieux et al., 2015b; Kwon et al., 2016; van Deursen et al., 2015), in its 2015 report, WHO had already suggested that much of the results concerning the negative consequences of addiction to smartphones and other electronic devices on well-being were possibly biased due to the lack of studies incorporating the time dimension (i.e., follow-up studies). This part of our study confirms that, as suggested by other researchers, the smartphone addiction has a negative effect on social support over time.
Moreover, results from latent growth mixture models indicate that smartphone addiction is associated with a decline in social support in the following 12 months and that this decrease is greater at higher initial smartphone addiction levels. These results complement those obtained in the cross-lagged causal models. Because of the negative relationship between social support and smartphone addiction, this decline in support is likely to be linked to higher rates of addiction in the future. This relationship, however, is hypothetical and other research should verify this assumption.
The results of the MANOVA and univariate ANOVA also allowed us to better identify the relationship between these trajectories of social support, personality, psychological distress, and addition to smartphone over time. These trajectories are not affected by sociodemographic variables and, therefore, the previous levels of addiction to the smartphone are key to understand the evolution of the social support in these two groups.
The group of users with a greater addiction to the smartphone and a more prominent reduction in social support over time scored higher on extroversion, neuroticism, and sensation seeking. Past research has linked extroversion and neuroticism with both substance abuse (Mccormick, Dowd, Quirk, & Zegarra, 1998) and behavioral addictions including Internet, mobile phones, and smartphones (Andreassen et al., 2013; Butt & Phillips, 2008; Tosun & Lajunen, 2010). Also, sensation seeking has been regarded as an antecedent of substance abuse (see Drane, Modecki, & Barber, 2017; Doumas, Miller, & Esp, 2017) and behavioral addiction in a number of researches (Billieux, 2012; Elhai et al., 2017; Herrero et al., 2017a). Our work adds to previous literature the empirical evidence that these relationships also occur over time, thus supporting the idea that stable aspects of personality could exert an important influence on the subsequent addiction to the smartphone with consequences on the psychosocial well-being of users.
As for the third of our hypotheses, those participants with a higher level of smartphone addiction in Wave 1 and a more pronounced decrease in social support during the 12 months of follow-up showed the higher levels of psychological distress (depressive mood) and smartphone addiction in Wave 3. This finding suggests that decreasing social support over time associated to smartphone addiction tends to produce a greater addiction to the smartphone and higher levels of psychological distress, affecting in a negative way the psychological well-being of the user.
Strengths and Limitations
One of the main strengths of the present study is its longitudinal nature. The absence of longitudinal studies on the antecedents and consequences of the addiction to smartphone has frequently been pointed out as a notable limitation in the scientific literature in this area (Herrero et al., 2017a; WHO, 2015). The present study is one of the first to use a longitudinal methodology to disentangle the relationships between smartphone addiction and the psychosocial well-being of users. Also, participants of the study were obtained from a representative sample of Internet users of Spain, which allows for greater generalizability of the results compared to other studies using convenient samples (i.e., university students). The analytical strategy implemented, from our point of view, has allowed us to explore the relationship between smartphone addiction and psychosocial well-being in a number of ways. This could also be considered an important strength of the study. On the one hand, it has allowed us to study their reciprocal relationships over time (cross-lagged model). On the other hand, it has allowed us to identify different trajectories among users (latent growth mixture models) as well as to characterize these trajectories regarding the personality and the psychological distress of the users (MANOVA and univariate analysis of variance).
Among the potential limitations of the study, the fact that 63% of Wave 1 participants also provided full data on Waves 2 and 3 might compromise the generalizability of study findings. Participants who left the study showed sociodemographic characteristics similar to those that remained in it—with the exception of age. This might suggest that dropping out of the study occurred almost randomly in terms of the sociodemographic characteristics of the participants. More important with regard to the observed relationships among the variables of the study is the fact that the permanence in the study was related to greater social support and lower addiction to the smartphone at Wave 1. It is not clear to what extent this self-selection of participants could have affected the results of the study. The fact that study participants had higher social support rates and a lower level of smartphone addiction in Wave 1 could have affected the magnitude of the statistical relationships observed in the present study. Actually, the observed relationship between social support and smartphone addiction at Wave 1 was lower (around a 35% lower) for the group with less variability in social support and smartphone addiction (the nondropouts group). From this point of view, in the absence of such self-selection for the lower levels of support and higher levels of smartphone addiction, the magnitude of the relationships observed in the present study may have been greater. Further research should provide more information to clarify this. Another potential limitation is the fact that the social support measure in our study does not allow to distinguish between online and offline social support. In this sense, smartphone-addicted users would have less social support either because they would be using their smartphones (including online social interaction) to compensate for the lack of real-life social interaction or because they engage in online/smartphone-based social interaction to boost their actual, real-world social capital and support (see Herrero et al., 2004; Kwak & Kim, 2017, for an analysis of online–off-line processes of support). Future research should be aimed to further disentangle the specific role of problematic smartphone use (and eventually addiction) in the maintaining of both off-line and online support.
Conclusions
The study of the addiction to the smartphone has increased substantially in the last years and has led to substantial knowledge of some of its antecedents and associated consequences. Despite this, it is worth noting that much of this scientific evidence comes primarily from correlational studies. Although such studies are ideal for tentatively contrasting relevant work hypotheses, they are not intended to analyze in depth the directionality of these hypotheses. From this point of view, our study allows us to identify some influences that also take into account the time dimension.
First, the results of the study make it possible to clarify the relationship between smartphone addiction and social support: Both are negatively affected over time. The greater the addiction, the less social support in the future, which in turn is associated with an increase in addiction again.
Second, those participants who experienced a decline in social support over time due probably to their addiction to the smartphone present a clear personality profile: They are more extroverted, score higher on neuroticism, and are more sensation seekers. Thus, they are more engaged with the external world and enjoy interacting with people (extroversion). Also, they present higher emotional instability and present problems to think clearly, make decisions, and cope effectively with stress (neuroticism). For these participants, there is a stronger drive for sensation seeking that lead them to pursue higher levels of stimulation in order to avoid perceiving the situations as unpleasant (sensation seekers). According to our data, this type of participant has greater chances of showing smartphone addiction and, probably as a consequence, a decline in social support over time.
Third, this decline in social support over time seems to be associated with increased smartphone addiction at the end of the 1-year period analyzed in this study as well as higher levels of psychological distress. Finally, a significant presence of response biases (i.e., social desirability) was not observed, which allows us to expect a greater generalizability of study results.
Overall, the present study has provided empirical evidence on the paradoxical potential effects of the smartphone on the psychosocial well-being of the person. For some study participants, excessive use of a social communication tool—the smartphone—led to addiction which ultimately negatively affected how socially connected they felt with people important to them.
The parallelism of this idea with that of Kraut et al. (1998) on the Internet paradox is intentional and, from our point of view, even more pertinent than 20 years ago. At present, people who physically share contexts for social interaction (i.e., social recreational activities such as meetings, meals, dinners, outings, etc.) often spend a part of that time connecting with others who are also likely to be physically sharing spaces of interaction with people with whom they partially stop interacting. And that does not bother anyone, because everyone is doing the same. Does this in any way affect those who engage in this type of behavior?
According to our results, a decrease in social support should be expected—feeling less connected with the important people in their lives—along with an increase in both smartphone addiction and psychological distress. And, it does not seem to be a process that characterizes specific social groups based on their sex, age, or educational level. It is a general process that could affect the entire population equally. Hence, it is of great relevance.
Footnotes
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.