Academic Achievement and Problem Behaviors among Asian Pacific Islander American Adolescents

The temporal order in the models is not intended to test causality. To examine whether GPA is a causal risk factor, it requires much more rigorous conditions than the current data allow, like randomized clinical trials (Kraemer, 2003).

There are some statistically significant differences between those who only participated in wave 1 and those who participated in both waves 1 and 2. These include the percentage of immigrants (χ ² = 32.64, p ≤ .000), whether mothers have received public assistance (χ ² = 3.92, p ≤ .05), and parental education (χ ² = −2.20, p ≤ .05). The samples that participated in both waves had a lower proportion of immigrants (8.49% vs. 11.11%), a higher proportion of maternal public assistance (10.64% vs. 9.64%), and a slightly higher parental education level (2.76 vs. 2.72). However, the results of the following analyses did not differ across these two groups. Wave 3 data are not used in this study because they were collected six years after the first interview, when respondents were no longer adolescents.

Multiracial youth are those who marked more than one race and ethnicity category on the survey. Recent studies show that multiracial youth as a group report significantly higher rates of problems (Choi et al., 2006; Udry et al., 2003). However, despite their rapidly growing population and potentially unique needs and heightened risk, multiracial adolescents are often not regarded as a distinct group in studies of adolescent problem behaviors and we know little about this group. Multiracial youth are, thus, included in this study as a separate group to provide baseline empirical findings, even if exploratory.

In wave 1, two additional items are available for aggressive delinquent offenses: frequency of serious physical fights and seriously injuring someone. When these two items are added to the scale, the alpha coefficient is 0.69 and the mean of the scale is 0.80 (SD = 1.21). The difference in internal consistency and the mean of the scale may be because the two items not assessed in wave 2 are more common among youth than the four items that were measured in both waves of interview. Analyses were conducted with the 6-item scale and the 4-item scale, respectively, in the models with cross-sectional data (i.e., using only wave 1 data). Despite the difference in alpha coefficients, the results were largely the same.

Although adolescents were also asked whether their parents have received any public assistance, the public assistance item was not used as a proxy for SES, because the reliability of adolescent responses on this item was not clear. For instance, the responses regarding father’s receiving assistance were largely missing. In addition, there was little difference in results when maternal public assistance was accounted for in the subsequent model testing.