Unconditional quantile regression has quickly become popular after being introduced by Firpo, Fortin, and Lemieux (2009, Econometrica 77: 953–973) and is easily implemented using the user-written command rifreg by the same authors. However, including high-dimensional fixed effects in rifreg is quite burdensome and sometimes even impossible. In this article, I show that when the number of fixed effects is large, the computational speed is massively increased by using xtreg rather than regress to fit the unconditional quantile regression models. I also introduce the xtrifreg command, which should be considered a supplement to rifreg. The xtrifreg command has many of the same features as rifreg but can be used to include a large number of fixed effects, to estimate cluster–robust standard errors, and to estimate cluster–bootstrapped standard errors.
4 July 2022 in Journal of Child Psychology and Psychiatry
Background: Children with ADHD tend to achieve less than their peers in school. It is unknown whether schools moderate this association. Nonrandom selection of children into schools related to variations in their ADHD risk poses a methodological problem. Methods: We linked data on ADHD symptoms of inattention and hyperactivity and parent–child ADHD polygenic scores (PGS) from the Norwegian Mother, Father, and Child Cohort Study (MoBa) to achievement in standardised tests and school identifiers. We estimated interactions of schools with individual differences between students in inattention, hyperactivity, and ADHD‐PGS using multilevel models with random slopes for ADHD effects on achievement over schools. In our PGS analyses, we adjust for parental selection of schools by adjusting for parental ADHD‐PGS (a within‐family PGS design). We then tested whether five school sociodemographic measures explained any interactions. Results: Analysis of up to 23,598 students attending 2,579 schools revealed interactions between school and ADHD effects on achievement. The variability between schools in the effects of inattention, hyperactivity and within‐family ADHD‐PGS on achievement was 0.08, 0.07 and 0.05 SDs, respectively. For example, the average effect of inattention on achievement was β = −0.23 ( SE = 0.009), but in 2.5% of schools with the weakest effects, the value was −0.07 or less. ADHD has a weaker effect on achievement in higher‐performing schools. Schools make more of a difference to the achievements of students with higher levels of ADHD, explaining over four times as much variance in achievement for those with high versus average inattention symptoms. School sociodemographic measures could not explain the ADHD‐by‐school interactions. Conclusions: Although ADHD symptoms and genetic risk tend to hinder achievement, schools where their effects are weaker do exist. Differences between schools in support for children with ADHD should be evened out.
A child’s environment is thought to be composed of different levels that interact with their individual genetic propensities. However, studies have not tested this theory comprehensively across multiple environmental levels. Here, we quantify the contributions of child, parent, school, neighbourhood, district, and municipality factors to achievement, and investigate interactions between polygenic indices for educational attainment (EA-PGI) and environmental levels. We link population-wide administrative data on children’s standardised test results, schools and residential identifiers to the Norwegian Mother, Father, and Child Cohort Study (MoBa), which includes >23,000 genotyped parent-child trios. We test for gene-environment interactions using multilevel models with interactions between EA-PGI and random effects for school and residential environments (thus remaining agnostic to specific features of environments). We use parent EA-PGI to control for gene-environment correlation. We found an interaction between students’ EA-PGI and schools suggesting compensation: higher-performing schools can raise overall achievement without leaving children with lower EA-PGI behind. Differences between schools matter more for students with lower EA-PGI, explaining 4 versus 2% of the variance in achievement for students 2 SD below versus 2 SD above the mean EA-PGI. Neighbourhood, district, and municipality variation contribute little to achievement (<2% of the variance collectively), and do not interact with children’s individual EA-PGI. Policy to reduce social inequality in achievement in Norway should focus on tackling unequal support across schools for children with difficulties.
The effectiveness of bullying prevention programs has led to expectations that these programs could have effects beyond their primary goals. By reducing the number of victims and perpetrators and the harm experienced by those affected, programs may have longer-term effects on individual school performance and prevent crime. In this paper, we use Norwegian register data to study the long-term impact of the Olweus Bullying Prevention Program (OBPP) on academic performance, high school dropout, and youth crime for the average student, which we call population-level effects. The OBPP program is widely acknowledged as one of the most successful programs reducing school-level bullying; yet, using a difference-in-difference design, no statistically significant population-level effects of the OBPP were found on any of the long-term outcomes in this study. When studied at the population level, as in the current project, the base rate prevalence of bullying is a major explanatory factor for these results. Earlier studies have shown that OBPP reduces bullying prevalence by 30–50%. This decrease translates into absolute reductions in bullying victimization and perpetration at the population level of “only” four and two percentage points, respectively. Our results suggest the average causal effects of school bullying involvement are too small to translate this reduction in bullying into a sizeable population-level impact on students’ long-term outcomes. However, a limited potential of anti-bullying programs to prevent population-level adversity can very well be compatible with substantial program effects for individual bullies and victims. Further, our results do not speak to the main objective of anti-bullying programs of limiting childhood abuse and safeguarding children’s human rights.
Schools and residential neighbourhoods constitute key contexts of development beyond the family of origin. Yet, few prior studies address whether the overall impact of these childhood contexts on adult life chances has changed over time. In this article, we investigate changes in socio-economic resemblance between former schoolmates and neighbouring children using Norwegian administrative data covering three decades. We use cross-classified multilevel models to decompose the variance in children’s educational attainment and adult earnings into the contributions found within and between their school and neighbourhood contexts in adolescence. We find that unadjusted school and neighbourhood correlations in educational attainment are relatively modest and declining over time. These trends largely reflect declining socio-economic segregation between schools and neighbourhoods over time. After adjusting for sorting by family background, schools account for 2 per cent or less of the total variation in completed years of education in the more recent cohorts and neighbourhoods even less. For adult earnings, the adjusted school correlations are very low, accounting for around 1 per cent of the total variance, while the contribution of neighbourhoods is close to zero. Our findings suggest that adolescent school and neighbourhood contexts are not major determinants of children’s later-life socio-economic attainments in the Norwegian welfare state setting.