Participants

A total of 241 families participated in the study. A forth (26.6%) were fathers and 177 were mothers (73.4%). The mean age of the participants was 39.4 years (SD = 6.84). Most were Spanish nationals (95.8%) and married (72.5%). Half (49.8%) were university graduates or equivalent, 31.5% had vocational training, and most (78%) were employed. Just over half of the children (n = 124; 51.7%) were girls, 116 (48.3%) were boys. They attended early childhood education (45%) or the first cycle of primary school (55%) in schools in the province of A Coruña (76.7% state funded, 18.8% independent, and 4.6% private).

Instrument

We used a questionnaire designed for this study about childhood mobile device use. It was made up of eight sections (sociodemographic information about the family and their children, identifying data about the use of mobile devices, time online, activity, use habits, parental control, and parental assessment of risks and benefits). During the design of the questionnaire, the content was validated by a panel of eight experts in methodology and educational technology from four countries (Brazil, Costa Rica, Ecuador, and Spain) who analyzed the unambiguity, relevance, and importance of each item. Their contributions helped us to improve the wording for various items and to restructure the initially proposed blocks.

The present study focused on analysis of the section related to Benefits (7 items) and Risks (8 items) of mobile device use, rated on a 5-point Likert scale (1 = Never and 5 = Always) which demonstrated excellent reliability (α = .885).

Construct validity was assessed using Exploratory Factorial Analysis (EFA) based on parallel analysis, a technique that uses the original correlation matrix, allowing common factors to be identified with eigenvalues that are greater than what would be obtained at random (^{Lloret-Segura et al., 2014}). The extraction method was Principal Axes, as the distribution did not exhibit normality (p < .001), with Oblimin oblique rotation, based on the intercorrelation between factors, have been shown to be one of the most appropriate modern options in social sciences (^{Lloret-Segura et al., 2014}; ^{López-Aguado & Gutiérrez-Provecho, 2019}).

After confirming that the structure of the data was suitable for EFA (KMO = .81; Bartlett's Sphericity: χ ² ₁₃₆ = 874.08, p < .001), we examined the factorial structure. Table 1 shows the solution extracted in two factors explaining 58.3% of the total variance (38.39% explained by Factor 1 and 19.41% explained by Factor 2). The results are presented ordered according to the pattern matrix, which shows the unique factor loadings on the variables, with loadings above .65 for Factor 1 and .5 for Factor 2. The structure matrix is also shown with the (zero order) correlations between factors and variables. There were no notable discrepancies between the two matrices. Loadings below .4 were removed, and there was no need to remove any items.

Procedure and data analysis

The questionnaire was applied during three months when the families came to the schools to drop off or collect their children. The aims of the study were explained to them, and they were assured of confidentiality and ethical guarantees, and that the data would only be used for the purposes of this study. After the explanation, and after any questions they had were answered, they were given a QR code with which they could download and complete the questionnaire on their own mobile phones.

We performed descriptive (central tendency and deviation) and inferential analyses. We used the t test (paired samples) to identify the families' ratings of the perceived risks and benefits of their children's mobile phone use. In addition, we examined the correlations between risks and benefits through Pearson's r. This test also allowed us to test the relationships and differences between benefits and risks according to time spent using the device and parental control. For the comparative analysis by educational stage, we used the t test (independent samples). Finally, we performed a Multivariate Analysis of Variance (MANOVA) to determine whether there were differences between the various benefits and risks and -with each acting as dependent variables- with respect to time spent using the device and parental control. We considered differences statistically significant at a 95% confidence level (p < .05).

Descriptive analysis of the risks and benefits of using mobile phones

In general, the families rated the risks higher than the benefits (Table 2). The paired-samples t test showed that most of these comparisons were statistically significant (p < .001), with a large effect size for the comparisons between all of the risk variables and the benefit of “Parental control” (d = .64-.95), with the exception of “Economic costs”, which had a small effect (d = .36). There were significant differences in the case of “More access to information” and the risk variables (p < .001) with a small effect size (d = .22-.38) in “Distorts reality” (t ₂₃₅ = 2.13, p = .04, d = .14), and in “Access to online games” (t ₂₃₅ = 2.43, p = .002, d = .16). “Economic costs” were rated significantly lower than “Better communication” (t ₂₃₅ = -2.82, p = .005, d = .62) and were not significantly different from the ratings for the benefit, “Useful resources” (t ₂₃₅ = 1.53, p = .127), although the other risks were, with small to moderate effects (p < .001, d = .25-.53). “Educational utility” and “Entertainment” were rated significantly below almost all of the risk variables (p < .001) except “Distorts reality” (R7: t ₂₃₅ = 1.59, p = .114; R8: t ₂₃₅ = 1.59, p = .113) and “Access to online games” (R7: t ₂₃₅ = 1.96, p = .051; R8: t₂₃₅ = 1.89, p = .06).

Looking at the Benefits factor, the parental control element was rated significantly higher than having “Better communication” (t ₂₃₅ = 6.24, p < .001, d = .41) and “Socialization” (t ₂₃₅ = 1.49, p < .001, d = .1), whereas it was rated significantly lower than the benefit of “More access to information” (t ₂₃₅ = -9.57, p < .001, d = .62), “Educational utility” (t ₂₃₅ = -10.57, p < .001, d = .69), the possibility of using “Useful resources” (t ₂₃₅ = -7.685, p < .001, d = .5), and even “Entertainment” (t₂₃₅ = -10.143, p < .001, d = .66).

The respondents gave the lowest ratings to using mobile devices for “Better communication”, which had a mean score below two points with a dispersion of responses of little more than one point. This item was rated significantly lower than all of the others in the factor (p < .001). “Educational utility” and “Entertainment” were rated significantly higher, followed by “More access to information”. In fact, “Educational utility” was rated significantly higher than “Socialization” (t ₂₃₅ = 14.68, p < .001, d = .95) and the use of “Useful resources” (t ₂₃₅ = 2.89, p < .001, d = .19). “Entertainment” was rated significantly higher than “Educational utility” (t ₂₃₅ = 0.11, p < .001, d = .001), “Socialization” (t ₂₃₅ = 15.01, p < .001, d = .98), and the use of “Useful resources” (t ₂₃₅ = 2.75, p < .001, d = .18), whereas “Better access to information” was rated significantly higher than “Socialization” (t ₂₃₅ = 12.63, p < .001, d = .82) and the use of “Useful resources” (t ₂₃₅ = 2.01, p < .001, d = .13). The effect size between the items referring to educational utility, entertainment, and access to information was small, while for the others it was moderate to large.

In the Risk factors, there were statistically significant differences, with families placing “Access to inappropriate content” as a greater risk than “Reduces performance” (t ₂₃₇ = 5.38, p < .001, d = .35), “Family and social isolation” (t ₂₃₇ = 5.88, p < .001, d = .38), “Addiction” (t ₂₃₇ = 3.09, p = .002, d = .2), “Economic costs” (t ₂₃₅ = 12.54, p < .001, d = .82), “Distorts reality” (t ₂₃₈ = 6.78, p < .001, d = .44), and “Access to online games” (t ₂₃₇ = 6.22, p < .001, d = .66). The second-highest rated risk was “Commit or suffer from cybercrime”, which was rated significantly higher than “Reduces performance” (t ₂₃₇ = 2.81, p = .005, d = .18), “Family and social isolation” (t ₂₃₇ = 4.66, p < .001, d = .3), “Economic costs” (t ₂₃₅ = 12.24, p < .001, d = .8), “Distorts reality” (t ₂₃₈ = 6.72, p < .001, d = .44), and “Access to online games” (t ₂₃₇ = 5.95, p < .001, d = .39). Similarly, “Addiction” was significantly more of a concern than “Family and social isolation” (t ₂₃₇ = 3.33, p < .001, d = .22), “Economic costs” (t ₂₃₅ = 11.3, p < .001, d = .74), “Distorts reality” (t ₂₃₈ = 5.27, p < .001, d = .34), and “Access to online games” (t ₂₃₇ = 4.11, p < .001, d = .51). Looking at the last three risks -rated significantly lower than the others (p < .001)-, “Access to online games” was significantly more of a concern than “Economic costs” (t ₂₃₄ = 8.11, p < .001, d = .53), but was significantly no different to “Distorts reality” (p = .56), which was rated significantly higher than the concern for “Economic costs” (t ₂₃₅ = 7.9, p < .001, d = .51).

Finally, Table 3 shows that there were weak, positive, significant correlations between some of the risks and benefits families considered about using mobile devices. In particular, families who gave a high rating to the use of mobiles to improve communication and as offering useful resources also thought that it could lead to lower performance, access to inappropriate content, family and social isolation, addiction, committing or suffering from cybercrime, and distortion of reality. In addition, they felt that the benefit of offering useful resources was associated with the economic costs of use.

Comparative analysis of the risks and benefits of mobile device use as perceived by families of children in early childhood and primary education.

Table 4 shows that, generally, the perceptions about benefits were slightly better in families of students in early childhood education than those with children in primary school. These differences were only statistically significant in relation to the potential for “Better access to information” (p = .042), with a small effect size (d < 0.3). In both groups, facilitating “Better communication” was not highly rated, with a greater dispersion in primary school children's families (SD = 1.12) than those in early childhood education (SD = 1.05).

When it comes to the risks, the families of children in early childhood education generally demonstrated greater concern in all of the items. These differences were statistically significant in relation to “Reduces performance” (p = .48), “Access to inappropriate content” (p = .022), “Family and social isolation” (p < .001), “Addiction” (p = .007), and “Committing or suffering cybercrimes” (p = .046).

Examination of the relationships and differences between the families' perceived benefits and risks of mobile device use by children's and parents' ages, by time spent online, and by the exercise of parental control

Parents' ages had an impact on how they rated the benefits of mobile device use in terms of better “Parental control” (r = -.231, p > .001) and about the “Educational utility” of these devices (r = -.147, p = .025). The older the parent, the less they felt they could exercise such control, and the less educational utility they saw. We found no statistically significant correlations between children's age and the perceived benefits of using a mobile device. In terms of risks, we found that parents' age had no relationship to their ratings of these variables, whereas the children's age was related to the rating for risk of “Family and social isolation”, with less worry in the family as children were older (r = .129). These relationships were weak, meaning the results should be considered with caution.

Box's M test for equality of covariance matrices showed that they were equivalent in the four groups analyzed in terms of weekly time spent using the device (Box's M = 108.561, p = .085). The MANOVA demonstrated differences in the ratings given to benefits as a function of time spent using the device (Wilks' Lambda = 0.844, F(21, 638) = 1.85, p = .012), with a small effect size (η² =.057). More specifically, considering the values of the inter-subject effects test, time spent using the device each week had an effect on the ratings related to “Parental control”, “Better communication”, and “Entertainment”. Post hoc testing with Bonferroni's correction indicated that there was “Better parental control” (p = .024) and “Better communication” (p = .033) when the child used the device for less than one hour per day compared to no use, as well as “Better communication” when the time spent using the device was more than one hour per day compared to sporadic use (p = .04). Using the device for “Entertainment” was considered more beneficial when the it was used for less than one hour per day compared to no use (p = .016) or sporadic use (p = .035).

In the case of weekend device use, the matrices were not equivalent in the test groups (Box's M = 166.79, p > .001), meaning we used Pillai's trace to analyze the multivariate significance of the main effects. The results were significant, with a small effect size (Pillai's trace = .25, F(21, 654) = 2.77; p < .001; η² = .091). The inter-subject effects test showed the presence of an effect of weekend device use on all of the benefits evaluated, except “Parental control”. More specifically, we found lower ratings in the no-use group than the one hour or more groups for “Better communication” (p = .017), “Better access to information” (p = .002), “Educational utility” (p < .001), “Socialization” (p = .018), and “Entertainment” (p < .001). In addition, there were worse ratings from those whose children did not use devices than those who used them for less than one hour in the whole weekend in “Better access to information” (p = .006), “Educational utility” (p < .001), “Socialization” (p = .027), and “Entertainment” (p = .034).

Looking at the risks according to time spent on the mobile device per week, Box's M test showed the absence of equality of the covariance matrices in the groups analyzed (Box's M = 222.97, p < .001), hence the data were examined using Pillai's trace (value = 0.108, F(24, 669) = 1.04, p < .405), which indicated that time of use had no significant effect on the ratings of risks. Looking at weekend use, it was also not possible to verify equality of the covariance matrices (Box's M = 337.87, p < .001). In this case, Pillai's trace showed that the results were significant, with a small effect size (Pillai's trace = .185, F(24, 651) = 1.78; p = .013; η² = .066). The inter-subject effects test showed that the effect of time spent on the mobile device on the risks referred solely to “Access to online games”, with this being rated higher when devices were used for less than one hour over the whole weekend than when the use was sporadic (p = .05) (Table 5).

Finally, we examined the effect of exercising parental control on the perceived benefits and risks. Looking at the benefits, the absence of equality in the matrix of covariances (Box's M = 92.18, p = .022) meant we used Pillai's trace to test the information (Pillai = 0.112, F(14, 444) = 1.87, p = .027, η² = .06), which showed that there was a statistically significant effect of parental control on the perception of benefits, albeit with a small effect. More specifically, the effect was seen in relation to the benefits “Parental control” (p = .01), “Better access to information” (p < .001), and “Educational utility” (p = .003). In each case the ratings were higher in families who reported exercising parental control that children complied with.

Looking at the risks, we again used Pillai's trace due to not finding equality in the covariance matrix (Box's M = 184.55, p < .001). This indicated the presence of a statistically significant effect for parental control on the perception of risks (Pillai = 0.161, F(16, 442) = 2.42, p = .002). The inter-subject effects test showed that these effects occurred in the risks associated with “Reduces performance” (p = .005), “Commit or suffer from cybercrime” (p = .025), “Addiction” (p = .006), “Economic costs” (p = .003), and “Access to online games” (p = .015). In each case, the parents gave a higher rating in families who exercised parental control than those who did not (Table 6).