Demographic Factors Affecting Children's Smoking Behavior

Essential in studying correlational relationships is accounting for covariates that control for the effects of other confounding factors on a child’s smoking behavior. In this subsection, we discuss the independent variables fitted in our regression models.
Children’s Demographic Characteristics We account for the children’s various demographic characteristics, namely, their age, level of education and sex. We hold for the ages and the squared ages of the children, as age can be indicative of the children’s independence with regards to decision-making, especially in deciding whether to smoke or not. Children’s education may be indicative of their knowledge on the health hazards associated with smoking behaviors. Knowing the risks

Unfortunately, the Cebu Longitudinal Health and Nutritional Survey does not report on total income of the children’s households for all the tracking years. In order to compensate, we used principal component analysis (PCA) in computing for the samples’ wealth index as a proxy for socioeconomic status (SES). PCA is a multivariate statistical analysis used primarily for the reduction of multiple related variables, in order to simplify and to make them easier to interpret (Lehe and Powell, 2010). We used asset-based measures which utilize information on household’s possession of specific durable assets, access to utilities and household characteristics as indicators for SES (see Table 6.1). This method is also employed by the Demographic and Health Surveys (The Demographic and Health Surveys, n.d.). In order to address truncation, where not enough variations in the variables is observed, we use 42 binary variables from the household data provided by CLHNS from the years 2002 and 2005, which is above the standard of at least twenty variables. We arrived with five quintiles for socio-economic status represented by following classifications: poorest (n=535), poorer (n=534), middle (n=535), richer (n=534) and richest (n=534) with the lowest factor scores to the highest factor scores, respectively. The importance for accounting for

We also regressed the calculated wealth index with the educational attainment of the children. Results were in line with common economic theories such as the wealth effect and Keynes’ Theory of Consumption that wealth is positively correlated with total expenditure and educational attainment of individuals. These prove the reliability of our measure. Wealth index is also preferred over using total expenditure as a measure for socioeconomic status because consumption tends to be more volatile, characterized by seasonality and trend (Kumaranayake et. al., 2006). Dates of interview vary from year to year, and from individual to individual, which may ultimately affect comparability of a SES measure based on expenditures. (1) Regression results are presented in