Homemade citizens: The development of political interest during adolescence and young adulthood

Abstract

Despite being among the most important indicators of political participation, relatively little is known about the origins and the development of political interest over the lifespan. The formative years between childhood and adulthood are generally considered a crucial phase in which future electors form and strengthen political habits. The aim of this research is to better understand this important stage by examining the way in which parental socialization and life-cycle events affect the formation and growth of political interest during adolescence and young adulthood. While parental influences are expected to take place during childhood and persist over-time, life-cycle events are considered to influence development in early adulthood for those adolescents who did not grow up in a highly politicized environment. We assess these assumptions by applying latent growth curve modeling and using the German Socio-Economic Panel, which spans from 1984 to 2007. Our findings confirm strong parental socialization effects on interest levels during teenage years. While life-cycle events are not found to strongly affect the development of political interest during the formative years, the transition to adulthood is indeed a more critical period for those individuals who did not acquire high levels of interest from their family.

Appendices

Appendix A

Table A1

Table A1 Summary statistics

Appendix B

Latent Growth Curve Models

LGC models are models with random slopes and random intercepts that permit each case in the sample to have a different trajectory over time or as in our case over age (Bollen and Curran 2006, p. xi). LGC models posit the existence of continuous underlying or latent trajectories. The actual scores of the movement on the dependent variable over the lifespan are not so much of interest. We use these repeated individual observations y _ia to estimate an underlying trajectory or line that best describes this growth, in our case of political interest. The following is the trajectory equation for such an unconditional LGC model, which does not consider covariates that affect the latent trajectories (B.1):

where y _ia is the observed value of the variable y, in our case political interest, for the ith case at age a, α _i is the random intercept, representing the initial status of y for case i. As a constant it is usually fixed to 1. β _i is the slope for individual i, measuring the ‘true’ rate of change for each individual as they grow older. λ _a is a vector of constants that measures the functional form of the time or aging process. In our case, we assume a linear increase in political interest while making a transition from the first to the second stage of the life cycle between the age of 17 and 35. ɛ _ia represents the stochastic error, based on the usual statistical assumptions. We assume that the mean disturbance is zero for all i and a, that the random intercepts and slopes are uncorrelated with the equation disturbance. Lastly, disturbances for different individuals should be uncorrelated.

The uniqueness of the model is in the precise consideration of each single case. Normal regression methods average the effects across individuals. LGC models on the other hand explicitly allow citizens to differ and, hence, are very suitable to account for unobserved individual heterogeneity. Therefore the most important components of the model are the random intercept and slope, which can be expressed by the following intercept (2) and slope equations (B.3) for the unconditional LGC model:

where μ _α and μ _β are the mean intercept and mean slope across all cases. The intercept equations (B.2) represents the individual intercept α _i as a function of the mean of the intercepts for all cases α _i and a disturbance ς _αi . Similarly, the slope equation (B.3) treats the individual slope β _i as a function of the mean of the slopes for all cases μ _β and a disturbance ς _βi .

We can combine the trajectory, intercept and slope equation into a single equation by substituting the right-hand sides of the intercept equations (B.2) and slope equations (B.3) for α _i and β _i respectively in the trajectory equations (B.1). The overall unconditional LGC model can hence be expressed as:

This is the combined model, in that it combines the three previous equations into a single equation. Often, the first term in the parentheses in equations (B.4) is referred to as the fixed component, which represents the mean structure of the general trend across all individuals. The second term on the other hand is described as the random component. This part of the model represents various sources of individual variability, intraindividual and interindividual differences from the overall mean trajectory. Remember that this model is unconditional in that the intercept and the slope equations only have the mean intercept and mean slope as determinants. Conditional LGC models simply include covariates that affect the random slope and intercept. An example of such a model that accounts for a covariate x₁, is as follows (B.5):