The mediating role of social avoidance between relative deprivation and smartphone addiction: a random intercept cross-lagged panel model study

Abstract

Background

Smartphone addiction represents a significant and growing public health challenge among adolescents. This longitudinal study examines the mediating role of social avoidance in the link between relative deprivation and smartphone addiction among Chinese junior high school students.

Methods

Data were collected from 639 participants over three waves at four-month intervals. By using a random intercept cross-lagged panel model (RI-CLPM), we were able to separate between-person and within-person effects.

Results

We found significant bidirectional predictive relationships between relative deprivation and smartphone addiction at the within-person level. Additionally, social avoidance served as a mediator in the longitudinal pathway from relative deprivation to smartphone addiction. Relative deprivation at T1 predicted increased social avoidance at T2, which then predicted higher levels of smartphone addiction at T3.

Conclusions

These results emphasize social avoidance as a key behavioral mechanism linking perceived inequality (i.e., relative deprivation) to problematic smartphone use. The study provides valuable insights for targeted interventions aimed at reducing smartphone addiction by addressing social avoidance behaviors in adolescents experiencing relative deprivation.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40359-026-04220-2.

Introduction

Adolescent smartphone addiction is an emerging public-health concern [1], marked by craving, loss of control, and continued use despite negative consequences [2]. Its prevalence is high among adolescents, who combine intensive smartphone engagement with ongoing neurocognitive maturation [3, 4]. Problematic use associates with diverse physical, psychological, and social harms, including insomnia, depression, anxiety, and social withdrawal [5–8]. Elucidating the developmental pathways that contribute to smartphone addiction is therefore essential for designing timely prevention and intervention.

Problematic mobile phone use is especially pronounced in macro‑social contexts marked by intense competition, such as the “involution” phenomenon in Chinese society. Junior high school students, whose self‑identity is still forming and whose cognitive control is not yet mature [9], often engage in social comparison in these environments [10]. These comparisons may produce a perceived disadvantage relative to peers and thus generate a subjective sense of relative deprivation [11]. Relative deprivation, defined as the negative cognitive and affective response to perceived injustice in social comparisons [12], has been identified as a key antecedent of maladaptive outcomes, including smartphone addiction [13]. To explain how this macro-level stressor is translated into individual addictive behavior, this study integrates Bronfenbrenner’s ecological systems theory with self-determination theory. Bronfenbrenner emphasizes that development occurs within nested environmental systems, in which the macrosystem—comprising cultural values and the competitive environment—exerts a distal yet significant influence [14, 15]. Self-determination theory provides a close-up psychological explanation: when the basic needs for autonomy, competence, and relatedness are frustrated, well-being is weakened, and compensatory behaviors emerge [16, 17]. We hypothesize that relative deprivation, as a macro-level stressor internalized by the individual, tends to frustrate these needs and may foster social avoidance as a maladaptive coping strategy, which could subsequently contribute to smartphone addiction.

Research on relative deprivation, social avoidance, and smartphone addiction

A growing body of research links relative deprivation to smartphone addiction. Longitudinal data show that relative deprivation robustly predicts later smartphone addiction [18, 19], and the relationship may be reciprocal, with addiction intensifying feelings of deprivation [19, 20]. Individuals prone to upward comparison experience greater relative deprivation, which, in turn, predicts problematic smartphone use [21, 22]. Although smartphone use can provide brief virtual relief, it undermines real-world functioning and reinforces negative self-perceptions, thereby deepening the deprivation it temporarily alleviates [23, 24].

Social avoidance plays a pivotal but underexplored role in this dynamic [25]. Adolescent social development depends heavily on positive peer interactions [26]. High levels of social avoidance correlate strongly with loneliness and reduced security and directly predict problematic smartphone use [27–29]. A bidirectional, reinforcing relationship likely exists in which smartphone addiction further impairs interpersonal skills and increases avoidance [30, 31]. Empirical evidence indicates that relative deprivation can promote social avoidance through two routes: by motivating withdrawal to escape negative social evaluation [32, 33] and by eliciting negative emotions that foster a search for online relief [34]. This avoidance redirects social interaction to the virtual sphere, providing short-term relief but gradually undermining real-world social skills, thereby creating a self-reinforcing cycle of greater virtual dependence and further withdrawal [35]. Neuroimaging data provide biological plausibility: experiences of deprivation and social rejection engage brain regions (e.g., the anterior insula) implicated in physical, which may amplify avoidance motivation [36, 37]. Despite these convergent findings, a key theoretical and empirical gap persists. Although each link has been documented, no longitudinal study has yet tested the complete mediating pathway proposed by our integrated model—specifically, whether social avoidance is the behavioral mechanism that converts experiences of relative deprivation (and the accompanying need frustration) into subsequent smartphone addiction over time.

Advantages of longitudinal research

Cross-sectional designs, while valuable for establishing correlations, cannot illuminate temporal precedence or dynamic processes [38, 39]. Longitudinal studies, by collecting time-series data, are essential for examining reciprocal relationships and temporal mediation. The RI-CLPM proposed by Hamaker et al. (2015) addresses a key limitation of the traditional Cross-Lagged Panel Model (CLPM) [40]: its difficulty in distinguishing between observed longitudinal associations and actual causal effects between variables or stable individual differences. The RI-CLPM separates longitudinal associations into between-person effects, which reflect correlations among stable trait components, and within-person effects, which reflect relationships among time-varying state components [41]. Between-person effects describe enduring individual differences in a variable’s average level across all time points. For example, individuals who, on average, score higher on smartphone addiction also tend to score higher on relative deprivation across the study period. Within-person effects describe how deviations from an individual’s own average at one time point relate to deviations at later time points. For example, if an individual’s smartphone addiction is higher than their own average at time T, does that predict higher or lower relative deprivation for that same individual at time T + 1? This provides insight into dynamic, time-specific processes within individuals [39, 42]. Given that prior research on these constructs has not used the RI-CLPM and interpretations from CLPMs involve uncertainty [19], employing the RI-CLPM offers a more rigorous method for exploring the dynamic interplay and mediation among relative deprivation, social avoidance, and smartphone addiction.

The current study

Guided by an integrated ecological systems and self-determination theory, this three-wave longitudinal study uses the RI-CLPM to examine dynamic relations among relative deprivation, social avoidance, and smartphone addiction in junior high school students. Our aims are twofold. First, we test reciprocal within-person associations between relative deprivation and smartphone addiction over time. Second, central to our model, we evaluate whether social avoidance mediates the within-person longitudinal pathway from relative deprivation to smartphone addiction. We hypothesize that:

• H1: Relative deprivation and smartphone addiction will show positive reciprocal cross-lagged effects at the within-person level.

• H2: The within-person impact of relative deprivation on subsequent smartphone addiction is hypothesized to be mediated by social avoidance.

This study aims to provide a detailed analysis of how competitive societal settings contribute to the development of smartphone addiction in a crucial developmental phase by examining macro-social stress, frustration of psychological needs, and compensatory actions in a longitudinal framework. The findings have the potential to advance integrated theory and inform prevention strategies that target adolescents’ core psychological needs and coping behaviors in high-pressure digital environments.

Method

Participants

The study used cluster sampling and selected three public ordinary junior high schools in Nanchang City, Jiangxi Province, China as sampling units. First, districts and counties within Nanchang City were sampled according to school size and urban–rural distribution. Then, in each selected school, whole classes were sampled by grade, and all in-school classes from grades 7 to 9 were included.

During the 2023–2024 academic year, all enrolled students in grades 7–9 at the selected schools were invited to participate. Inclusion criteria were: ① official student status at the school; ② age between 12 and 15 years; ③ informed consent from both the students and their legal guardian. Exclusion criteria were: ① severe cognitive or psychiatric disorders preventing completion of the questionnaire; ② long-term leave or suspension during the survey period; ③ explicit refusal by the guardian to participate.

Data were collected at three time points: September 2023 (T1, baseline), January 2024 (T2), and May 2024 (T3), with roughly four months between waves. At baseline (T1), informed consent forms were distributed to 892 students and their legal guardians, and 806 students completed the assessment (response rate: 90.4%). At T2, 746 students participated (attrition from T1 to T2: 7.4%), and at T3, 685 students participated (attrition from T2 to T3: 8.2%). A total of 639 participants comprised the final valid sample after screening, with all participants completing all three surveys and having less than 5% missing data per wave.

The demographic characteristics of the 639 valid cases were as follows. There were 299 males (46.8%) and 340 females (53.2%), with a mean age of 12.78 years (SD = 0.97). By family residence, 416 students (65.10%) lived in urban areas, 85 students (13.30%) lived in rural areas, and 138 students (21.6%) did not report this information. For family structure, 114 students (17.84%) were only children, 386 students (60.41%) were non–only children, and 139 students (21.75%) did not report this status. Grade distribution was 213 students (33.33%) in seventh grade, 361 students (56.50%) in eighth grade, and 65 students (10.17%) in ninth grade.

The primary causes of sample attrition were student absence due to illness, transfer to another school, or family relocation on the survey day. Attrition analysis revealed no significant differences between those who completed the study and dropouts in terms of gender (χ² (1) = 0.006, p > 0.05), age (t = 0.533, p > 0.05), T1 relative deprivation (t = 0.518, p > 0.05), T1 social avoidance (t = -0.645, p > 0.05), or T1 smartphone addiction (t = 0.767, p > 0.05), showing no systematic attrition. Little’s MCAR test confirmed that the missing data were random (χ² (141) = 132.547, p > 0.05). Consequently, Full Information Maximum Likelihood estimation with robust standard errors (FIML) was used to handle missing data.

Measures

Relative deprivation scale

Relative deprivation was assessed using the 4-item unidimensional Relative Deprivation Scale [43]. Participants rated items (e.g., “Compared to others, I feel I deserve more”) on a 6-point Likert scale (1 = strongly disagree, 6 = strongly agree), with higher mean scores indicating greater perceived relative deprivation. The scale demonstrated acceptable internal consistency across all waves: Cronbach’s α = 0.70 (T1), 0.77 (T2), and 0.81 (T3).

Social avoidance and distress scale

Social avoidance was measured using the 14-item social avoidance subscale of the Social Avoidance and Distress Scale [44]. Participants rated items (e.g., “I avoid social situations that force me to be with strangers”) on a 5-point Likert scale (0 = completely disagree, 4 = agree entirely), with higher average scores indicating greater social avoidance tendencies. The subscale showed excellent internal consistency across waves: Cronbach’s α = 0.84 (T1), 0.85 (T2), and 0.86 (T3).

The mobile phone addiction index

The Mobile Phone Addiction Index (MPAI) [45] was used to assess tendencies toward smartphone addiction. This 17-item scale, derived from DSM-IV addiction criteria, comprises four dimensions: loss of control (e.g., “Your friends and family complained about your time spent on your mobile phone”), withdrawal (e.g., “I felt restless and irritable if you had not used the mobile phone for some time.”), escape (e.g., “I use my smartphone to make myself feel better when I am in a bad mood”), and functional impairment (e.g., “My smartphone use has reduced my productivity”). Items are rated on a 5-point Likert scale (1 = strongly disagree to 5 = strongly agree), with higher mean scores indicating greater addiction severity. This study used the total score of the MPAI as the main analysis to provide general evidence of the association with overall smartphone addiction. Across the three waves, the total scale produced Cronbach’s α values of 0.93 (T1), 0.95 (T2), and 0.96 (T3). Subscale reliability at T1 was 0.89 (Loss of Control), 0.82 (Withdrawal), 0.90 (Escape), and 0.85 (Functional Impairment); at T2, 0.93 (Loss of Control), 0.91 (Withdrawal), 0.90 (Escape), and 0.86 (Functional Impairment); and at T3, 0.93 (Loss of Control), 0.92 (Withdrawal), 0.91 (Escape), and 0.90 (Functional Impairment).

Procedure

Data were collected using paper-and-pencil questionnaires in classrooms, with approval from school administrators and teachers. In addition to the main study variables, demographic information such as gender, age, and place of origin was also recorded. The surveys were conducted by undergraduate applied psychology students who received training for this task. They were assured of the confidentiality of their responses and that the data would be used solely for academic research, without any personal repercussions. Standardized procedures were consistently followed across all assessments, conducted every 4 months.

Statistical analyses

Epidata was used to screen and standardize responses from the returned questionnaires. Data were then processed in SPSS 26.0, where we performed pre-processing tasks such as variable transformation and scale reliability testing and conducted subsequent analyses, including descriptive statistics. Measurement invariance across three time points was evaluated with Mplus 8.0. After establishing longitudinal measurement invariance, we specified a Random Intercept Cross-Lagged Panel Model (RI-CLPM) in Mplus 8.0 to examine reciprocal relations and the mediating role of social avoidance over time.

The key model specifications were as follows. To account for shared occasion-specific variance, residual covariances among the three observed variables within each wave were freely estimated. No equality constraints were imposed on cross-lagged or autoregressive paths across the two-time intervals, allowing predictive relationships and stability effects to vary. Age (standardized) and gender (0 = male, 1 = female) were included as predictors of the random intercepts to control for the influence of these time-invariant covariates on individuals’ stable trait levels. The model was estimated via maximum likelihood (ML), chosen for its compatibility with the planned bias-corrected bootstrap test of mediation and because the distributions of the key continuous variables did not show substantial departures from normality. Finally, the longitudinal mediating effect of social avoidance on the relationship between relative deprivation and smartphone addiction was tested using a bias-corrected bootstrap with 5,000 resamples in Mplus 8.0.

Results

Common method bias

Self-reported measures of relative deprivation, social avoidance, and smartphone addiction can introduce common method bias. To address this, Harman’s single-factor test was conducted. The study’s longitudinal design enabled analysis at three distinct time points. Results showed that the number of factors with eigenvalues over 1 was 7, 6, and 5, respectively. The variance explained by the first factor was 30.49%, 36.29%, and 37.73%, all below the 40% threshold. Therefore, no significant method bias was detected in this study.

Descriptive and correlation analyses

Results showed that relative deprivation, social avoidance, and smartphone addiction had significant positive cross-time correlations at all three points (0.416 ≤ r ≤ 0.733, ps < 0.001), indicating moderate to high stability over time. Additionally, there were significant positive correlations within the same time point (0.338 ≤ r ≤ 0.520, ps < 0.001) and across different time points (0.239 ≤ r ≤ 0.519, ps < 0.001) among these variables (see Table 1 for detailed results).

Table 1.

The mean, standard deviation, and correlation analysis results for each variable

Variable	M	SD	1	2	3	4	5	6	7	8	9
1	2.251	0.970	1
2	2.233	1.023	0.501	1
3	2.360	1.020	0.416	0.510	1
4	1.693	0.691	0.332	0.272	0.284	1
5	1.668	0.688	0.404	0.341	0.355	0.682	1
6	1.711	0.692	0.303	0.257	0.357	0.624	0.731	1
7	2.024	0.780	0.359	0.308	0.274	0.460	0.434	0.373	1
8	2.117	0.854	0.384	0.432	0.338	0.479	0.519	0.472	0.684	1
9	2.232	0.862	0.239	0.284	0.361	0.480	0.520	0.511	0.548	0.634	1

1=T1 Relative deprivation, 2=T Relative deprivation, 3=T3 Relative deprivation, 4=T1 Social avoidance, 5=T2  Social avoidance, 6=T3 =Social avoidance, 7 =T1  Smartphone addiction, 8=T2 Smartphone addiction, 9=T3 Smartphone addiction. All correlations were significant at p < 0.001

Longitudinal measurement invariance

This study, conducted across three time points, first examined the longitudinal invariance of the scale to ensure it consistently measures a specific trait over time. In Mplus 8.0, items for each variable served as indicators of latent variables for relative deprivation, social avoidance, and smartphone addiction. Configural, metric, and scalar invariances were tested in sequence. Differences in fit indices under various restricted invariance conditions were then compared. Because the chi-square statistic is sensitive to sample size, changes in CFI (ΔCFI) and RMSEA (ΔRMSEA) were used to evaluate model fit changes. According to Chen(2007) [46], ΔCFI ≤ 0.01 and ΔRMSEA ≤ 0.015 indicate test equivalence across the three assessments for the same variable.

The study showed that relative deprivation, social avoidance, and smartphone addiction achieved full scalar longitudinal measurement invariance (see Table 2 for detailed results). This means these constructs meet the criteria for longitudinal measurement invariance, enabling meaningful comparisons across the three measures. As a result, the random-intercept cross-lagged panel model was appropriate for further analysis of the relationships among these variables.

Table 2.

Tests of measurement invariance

Variable	Model	χ²	df	AIC	CFI	TLI	RMSEA	SRMR	Model Comparison	ΔCFI	ΔRMSEA
Relative Deprivation	M1Configural Invariance	42.045	39	22973.607	0.999	0.998	0.011	0.027
	M2 Metric Invariance	51.253	45	22970.815	0.998	0.997	0.015	0.030	M2-M1	-0.001	0.004
	M3Scalar Invariance	63.540	51	22971.102	0.996	0.994	0.020	0.032	M3-M2	-0.002	0.005
Social Avoidance	M1Configural Invariance	1471.776	762	73603.227	0.943	0.935	0.038	0.049
	M2 Metric Invariance	1517.745	786	73601.196	0.941	0.935	0.038	0.051	M2-M1	-0.002	0.000
	M3Scalar Invariance	1564.383	810	73599.834	0.939	0.935	0.038	0.052	M3-M2	-0.002	0.000
Smartphone Addiction	M1Configural Invariance	3302.338	1107	72809.726	0.924	0.912	0.056	0.054
	M2 Metric Invariance	3357.014	1133	72812.402	0.923	0.913	0.055	0.055	M2-M1	-0.001	-0.001
	M3Scalar Invariance	3438.794	1159	72842.182	0.921	0.913	0.055	0.055	M3-M2	-0.002	0.000

^*p<0.05, ^**p<0.01, ^***p<0.001

The bidirectional relationship between relative deprivation and smartphone addiction

Figure 1 presents the RI-CLPM results for relative deprivation and smartphone addiction, which fit the model well: χ² = 24.945, df = 13, CFI = 0.991, TLI = 0.981, RMSEA = 0.038, SRMR = 0.026. At the between-person level, relative deprivation and smartphone addiction were positively correlated (r = 0.511, p < 0.001). Additionally, age showed a significant positive association with the random intercept of relative deprivation (β = 0.124, p = 0.011), indicating that older adolescents tended to report higher smartphone addiction. At the within-person level, the autoregressive paths for both constructs were significant. For cross-lagged effects, relative deprivation at T1 positively predicted smartphone addiction at T2 (β = 0.139, p = 0.009), and relative deprivation at T2 positively predicted smartphone addiction at T3 (β = 0.135, p = 0.008). Conversely, smartphone addiction at T1 positively predicted relative deprivation at T2 (β = 0.128, p = 0.016), and smartphone addiction at T2 positively predicted relative deprivation at T3 (β = 0.142, p = 0.014). These findings indicate a reciprocal predictive relationship between the two constructs at the within-person level.

Fig. 1.

Random intercept cross-lagged model between relative deprivation and smartphone addiction. Note: The coefficients in the figure are standardized path coefficients; solid lines represent significant path coefficients, and dashed lines represent non-significant path coefficients. RI represents the random intercept. *p < 0.05, **p < 0.01, ***p < 0.001

Mediating effect of social avoidance

The intraclass correlation coefficient (ICC) was calculated to assess the extent to which individual differences explain variation in each variable. ICC analysis revealed that individual fluctuations accounted for 47.6% of the variation in relative deprivation, 67.5% in social avoidance, and 60.3% in smartphone addiction. Therefore, using random-intercept cross-lag models is essential to capture stable personal traits over time and provide a more accurate understanding of the relationships among these variables.

A random intercept cross-lagged model was developed using structural equation modeling (SEM) to distinguish between individual-level and within-individual effects. This model was utilized to examine the relationships among relative deprivation, social avoidance, and smartphone addiction with greater precision (see Fig. 2). The results indicated a good model fit: χ² = 28.464, df = 15, CFI = 0.995, TLI = 0.981, RMSEA = 0.038, SRMR = 0.021.

Fig. 2.

Random intercept cross-lagged analysis of relative deprivation, social avoidance, and smartphone addiction. Note: Dashed paths indicate non-significant results. For clarity, correlation coefficients at different time points are not shown

The RI-CLPM results, detailed in Fig. 2, reveal significant positive correlations among the intercepts for relative deprivation, social avoidance, and smartphone addiction at the between-person level (β = 0.577, p < 0.001; β = 0.721, p < 0.001; β = 0.428, p < 0.001). This indicates that junior high school students experiencing higher levels of relative deprivation are more likely to exhibit social avoidance and smartphone addiction tendencies. Moreover, students with elevated social avoidance levels are also prone to higher rates of smartphone addiction. Additionally, the model accounted for the effects of covariates: gender significantly and positively predicted the random intercept for relative deprivation (β = 0.108, p = 0.02), and age significantly and positively predicted the random intercept for social avoidance and smartphone addiction (β = 0.102, p = 0.005, β = 0.145, p = 0.011). At the within-person level, autoregressive paths indicate that the coefficients for sense of relative deprivation across time points are all significant, reflecting high temporal stability. Smartphone addiction shows low stability from T2 to T3, but relatively high stability from T1 to T2. Likewise, social avoidance is unstable from T1 to T2 yet exhibits relatively high stability from T2 to T3. The cross-lagged path indicates that within-person fluctuations in relative deprivation at T1 significantly and positively predict within-person fluctuations in social avoidance at T2 (β = 0.271, p = 0.001). However, the predictive effects of within-person fluctuations in social avoidance on within-person fluctuations in relative deprivation were not significant. However, fluctuations in social avoidance did not predict changes in relative deprivation. A significant bidirectional positive predictive effect exists between relative deprivation and smartphone addiction from Time T1 to T2 (β = 0.263, p < 0.001; β = 0.197, p = 0.012), but smartphone addiction at T2 significantly predicts relative deprivation at T3 (β = 0.151, p = 0.045). Changes in social avoidance at T2 significantly predict changes in smartphone addiction at T3 (β = 0.190, p = 0.032), whereas smartphone addiction does not significantly predict fluctuations in social avoidance.

Based on this, using Mplus 8.0 and the Bootstrap method with 5000 resamples, the longitudinal mediating role of social avoidance in the relationship between relative deprivation and smartphone addiction was examined. The analysis showed that, after accounting for between-individual differences, social avoidance significantly mediated the link between relative deprivation and smartphone addiction. Specifically, the indirect effect from relative deprivation to smartphone addiction through social avoidance was significant (Indirect = 0.051, 95%CI = 0.002–0.101). Conversely, the path from smartphone addiction to relative deprivation via social avoidance was not significant.

Discussion

This study used a three-wave longitudinal design and the RI-CLPM statistical model to examine the dynamic relationship among relative deprivation, social avoidance, and smartphone addiction in Chinese junior high school students. At the within-person level, results supported a reciprocal relationship between relative deprivation and smartphone addiction (H1). More importantly, and central to our theoretical model, social avoidance was identified as a significant longitudinal mediator in the pathway from relative deprivation to subsequent smartphone addiction (H2). The findings indicate that adolescents with a higher sense of relative deprivation are more likely to engage in social avoidance behaviors, which then increases the risk of smartphone addiction.

Discussion of the relationship between relative deprivation and smartphone addiction

The results reveal a mutually reinforcing cycle between relative deprivation and smartphone addiction at the within-person level, supporting Hypothesis 1 and consistent with prior empirical work [19, 47, 48]. From an Ecological Systems Theory perspective, the macrosystem’s competitive climate and the prevalence of “involution” foster frequent upward social comparisons [49]. When those comparisons are internalized as relative deprivation, they frustrate basic psychological needs in ways predicted by Self-Determination Theory: autonomy declines as people feel controlled by external social standards; negative self-evaluations undermine competence; and relatedness is compromised by experiences of social exclusion or perceived inferiority [2]. Smartphones often serve as a compensatory outlet because they are accessible, anonymous, and deliver immediate virtual rewards [50]. Excessive use may provide brief feelings of control, simulated social connection, and escape from real-world stressors, thereby promoting addictive patterns of use [27, 51, 52]. In turn, such addictive behavior may impair everyday functioning, reinforce negative self-perceptions, and increase social alienation, thereby intensifying feelings of relative deprivation [53]. Longitudinal evidence indicates that relative deprivation not only increases vulnerability to addictive use but is also maintained or worsened by the social deficits that result from that use [19], creating a persistent negative feedback loop across psychological and behavioral domains.

Discussion of the mediating effect of social avoidance

A key contribution of this study is the longitudinal demonstration that social avoidance mediates the within-person link between relative deprivation and smartphone addiction, thereby supporting Hypothesis 2. According to ecological systems theory, this mediating pathway likewise illustrates how macro-level competitive pressures are transmitted through individual perceptions and behaviors to produce specific addictive outcomes. The RI-CLPM showed a bidirectional predictive relation at the within-person level. The results show that prior relative deprivation predicted subsequent increases in smartphone addiction, and prior addiction predicted later increases in perceived deprivation [32, 53]. This reciprocal effect was significant only between T1 and T2, subsequently weakening from T2 to T3. This finding suggests that, over time, individuals may deplete their psychological resources [54] due to ongoing relative deprivation and smartphone addiction, resulting in a notable bidirectional predictive effect. However, individuals may also replenish their psychological resources through various means, such as seeking social support or engaging in different activities, thereby diminishing this bidirectional predictive relationship. The reasons for this attenuation warrant further investigation; potential explanations include developmental adaptation, resource conservation, or the implementation of alternative coping strategies. Nonetheless, the observed pattern underscores the importance of early interventions that address perceived injustice and maladaptive coping before these cycles become entrenched.

Mediation analysis indicated that relative deprivation at T1 predicted greater smartphone addiction at T3 indirectly through increased social avoidance at T2. This result aligns with evidence that relative deprivation elicits negative affect and social withdrawal [54, 55]. As a result, individuals rely more on virtual interactions and immersive online activities for compensation [56], forming reinforced use patterns that foster later addictive behavior [57]. Notably, the cross-lagged mediation emerged for within-person fluctuations, showing that the process reflects dynamic within-person changes over time rather than stable between-person differences. Thus, when a person feels more deprived than usual at a given moment, the increase more strongly predicts their own subsequent rise in problematic phone use, rather than reflecting a general predisposition to addiction. This shows a positive feedback loop within individuals. A temporary surge in perceived deprivation leads to increased avoidance and compensatory smartphone usage, resulting in heightened dependency, which could further amplify future feelings of deprivation. The reverse indirect path from smartphone addiction to subsequent relative deprivation via social avoidance was not significant, which may reflect substantial within-person variability in avoidance or the action of other untested mediators (e.g., emotion-regulation strategies, perceived social support) [31, 58].

These findings advance both theoretical and practical understanding of adolescent smartphone addiction. They propose a sequential pathway from social deprivation to behavioral avoidance, leading to virtual compensation, which shows how persistent frustration of real‑world social needs can prompt avoidance and, in turn, maladaptive, immersive smartphone use as a compensatory strategy. The framework redirects attention from symptom‑level management (for example, mere screen‑time reduction) to the underlying motivational deficits and coping behaviors. Social avoidance thus becomes a primary target for intervention. Techniques such as cognitive restructuring, social skills training, and exposure therapy may provide informative targets for helping adolescents manage distress from social comparison and foster authentic face-to-face relationships. Reducing reliance on virtual compensation through these approaches could be especially relevant in collectivist cultural settings such as China. Finally, the use of RI‑CLPM confirms that these processes reflect active within‑individual dynamics rather than stable between‑person differences, underscoring the need for timely, personalized intervention strategies.

Mechanisms within the Chinese context and implications for intervention

The Chinese sociocultural context, marked by collectivism and a strong emphasis on academic achievement and social harmony, makes adolescents especially sensitive to relational dynamics and social comparison [59, 60]. In this environment, relative deprivation can exert a pronounced and specific effect on the need for relatedness [61]. The observation that social avoidance functions as a primary mediator highlights the importance of sustaining and repairing face-to-face social ties among Chinese youth. Preventing withdrawal from interpersonal interactions may therefore be important for interrupting the link between perceived inequality and digital addiction.

Our RI-CLPM analysis has concrete implications for intervention at two distinct levels. First, between-person differences show that some adolescents consistently report higher relative deprivation and greater smartphone addiction, which supports universal or selective prevention strategies emphasizing cognitive restructuring of social comparisons and resilience building. Second, the within-person dynamics indicate that episodic increases in perceived deprivation may precede social withdrawal and short-term elevations in addiction risk. This pattern underscores the potential value of timely, individualized interventions. Approaches such as social skills training, exposure-based methods, and the enhancement of adaptive coping strategies to manage social-comparison distress might be considered to reduce reliance on virtual forms of compensation.

Limitations and future directions

While the study offers valuable insights using the RI-CLPM model and a rigorous design, several limitations remain. First, although the self-reported data were tested for common method bias, they remain vulnerable to subjective bias. This issue arises from the difficulty of incorporating multidimensional objective measurement tools (e.g., mobile phone usage tracking, observations of socially evaluated behaviors) into longitudinal studies. Second, the study did not explore the moderating effects of contextual factors such as family functioning and school climate. Third, the focus on Chinese junior high school students may limit the cross-cultural applicability of the results, particularly given cultural differences in social comparison tendencies among Western adolescents [62]. Fourth, in this study, we used only the total score of the Mobile Phone Addiction Index Scale for analysis and did not examine whether specific dimensions (such as loss of control or withdrawal symptoms) show distinct associations with relative deprivation and social avoidance. Future studies that analyze factor scores will more precisely elucidate how different manifestation patterns of smartphone addiction relate to these constructs.

To address the remaining issues, several recommendations for future research are suggested. First, incorporating behavioral measures, such as digital phenotyping of smartphone use and experimental paradigms for social avoidance, can capture real-time dynamics and reduce reliance on self-reports. Second, developing moderated mediation models with contextual moderators (e.g., parenting style, school support) could clarify under what conditions the mediated pathway is either strengthened or weakened. Third, cross-cultural and cross-age group comparisons are crucial to evaluate the generalizability of the “social deprivation–avoidance–addiction” pathway, especially between individualistic and collectivistic societies. Ultimately, integrating neurophysiological measures (e.g., EEG, fMRI) may reveal the neural mechanisms underlying these behavioral relationships, providing a more comprehensive understanding of the etiology of smartphone addiction.

Conclusion

This longitudinal study, using a three-wave design and the RI-CLPM, demonstrated that relative deprivation among junior high school students positively predicts smartphone addiction via social avoidance. The findings confirmed a significant bidirectional predictive relationship between the constructs at the within-individual level. At the same time, the mediating role of social avoidance was observed only in the direction from relative deprivation to smartphone addiction. Notably, social avoidance mediated only the pathway from relative deprivation to smartphone addiction. This is the first longitudinal validation of social avoidance as a behavioral mediator, addressing the limitations of previous cross-sectional studies. Although there may be bias from self-reported data and the focus on Chinese adolescents, which could limit cross-cultural applicability, the findings highlight the potential importance of early identification and interventions targeting social avoidance in efforts to disrupt the addiction cycle. Future research could incorporate objective behavioral measures, examine contextual moderators, and perform cross-cultural comparisons to enhance the effectiveness of intervention strategies.

Authors’ contributions

HW designed the study, drafted the manuscript; ZZ. gathered the data, performed the analysis, and interpreted the data; YB. and DQ revised the manuscript and controlled the quality. All authors read and examined the final manuscript.

Funding

This study was supported by the National Natural Science Foundation of China (72164018), the Humanities and Social Sciences Research Planning Foundation of Ministry of Education (22YJA190012), the Jiangxi University Humanities and Social Sciences Research Project (Grant No. SZZX23019), the Project of the Jiangxi Ideological and Political Education Research Association for Higher Education Institutions (Grant No. XLJK24212).

Data availability

Due to privacy protection, the data and materials are not publicly available but can be provided upon reasonable request to qualified researchers.

Declarations

Ethics approval and consent to participate

All experimental protocols were approved by the ethics committee of the School of Psychology, Jiangxi Normal University. All methods were carried out in accordance with relevant guidelines and regulations. Informed consent was obtained from all participants.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.