Uyan-Semerci, P., Erdoğan, E., Akkan, B., Müderrisoğlu, S., & Karatay, A. (2017). Contextualizing subjective well-being of children in different domains: Does higher safety provide higher subjective well-being for child citizens? Children and Youth Services Review, 80, 52-62.
Children’s subjective well-being is determined by the political, social and economic conditions in their sociocultural environments. This study explored the relationship between safety welfare context of the country and children’s subjective well-being in the following six domains: health (e.g., satisfaction with your health), material (e.g., satisfaction with the house or flat where you live), education (e.g., satisfaction with your school experience), relationship (e.g., satisfaction with your family life), risk and safety (e.g., satisfaction with how safe you feel), and self-perception (e.g., satisfaction with your own body). The authors also examined whether this relationship was moderated by age and gender of the children.
This study used the second wave data collected from 15 countries (35,417 children aged from 10 to 12) in the International Survey of Children’s Well-Being to investigate how macro level safety welfare context was related to children’s subjective well-being in different domains. In order to classify different countries into high, medium, and low safety welfare contexts, macro level indicators (from the United Nations Programme on Development and The World Social Protection Report 2014 - 2015) of social policies in terms of health, education, social services, and societal safety and risk for children in these countries were used for hierarchical cluster analysis. After countries were grouped into high, medium and low safety welfare contexts, regression analysis was conducted to test the moderating roles of age and gender on the relationship between safety welfare context and children subjective well-being.
Results showed that there were significant variations between safety welfare contexts and children’s subjective well-being in different domains, and these variations differed according to gender and age. For instance, children reported higher subjective well-being in the domains of ‘material’ and ‘risk and safety’ among high and medium safety welfare contexts. Girls reported significantly lower subjective well-being than boys in the low safety welfare context. However, girls reported higher subjective well-being than boys in the domains of ‘education’ and ‘relationship’ across different safety welfare contexts. Furthermore, children from the low safety welfare context had the lowest average on subjective well-being in their `relationship’ domain, but older children reported higher satisfaction in this domain, which indicated that the gap across different safety welfare contexts diminished among older children. The findings of this paper illustrated that macro level social policies can affect individual child’s subjective well-being directly and indirectly in different domains. The impact of social policies may differ across gender and age groups.
The second wave of the International Survey of Children’s Well-being data consisted of nested data from 15 countries with at least 1000 children per country. Due to the small sample size at the higher level (i.e., 15 countries), it is understandable that the authors did not use multilevel modeling to analyze the data. Instead, they opted for cluster analysis to group countries together to create different levels of safety welfare contexts as predictors in their regression analysis. However, the authors noted that with cluster analysis, the membership of different countries in the three safety welfare contexts would be highly susceptible to the choices of the macro level indicators, as well as the number and the type of countries in the cluster analysis. In other words, with different macro level indicators and different countries, it may pose great challenges in future studies to replicate the results in this study. Potential solutions could be to use macro level indicators in each country directly as Level 2 predictors in multilevel modeling with Bayesian MCMC estimations or to apply Kenward-Rogers adjustment with restricted maximum likelihood estimations (McNeish & Stapleton, 2016).
References
McNeish, D. M., & Stapleton, L. M. (2016). The effect of small sample size on two-level model estimates: A review and illustration. Educational Psychology Review, 28(2), 295-314. Doi: 10.1007/s10648-014-9287-x