In model E in Table 3.3, we replace the school fixed effects with the average school performance. We find a positive association of this variable with students’ skill production throughout primary school. So, it appears that students with better average performance are also in schools where students progress more throughout primary education. This finding, however, could be explained by factors such as school segregation or composition. 3.5.2 Second-sixth-grade skill relationship, a comparison between OLS and IV estimates While the findings of Section 3.5.1 estimated by OLS are informative, they presumably suffer from attenuation bias caused by measurement error in the independent skill variable. Consequently, we might underestimate the relationship of this variable with the sixth-grade skill score and overestimate the relevance of other variables included in our model. Therefore, we perform similar analyses as in Table 3.3, using an IV approach. We present the estimates of these analyses in Table 3.4, in which we show the relationship between the skill levels of students in the middle of second grade and middle of sixth grade.14,15 We show, similar to Table 3.3, the estimates for models D and E. The former model includes school fixed-effects to account for school het14Note that we cannot make a similar analysis for the relationship between the firstand sixth-grade skill score as we do not have two observations for reading comprehension in first grade. 15Tables B.5 and B.6 investigate heterogeneity in these second-sixth-grade relationship at the student level (parental education, sex, and migration background) and at the school level (number of schools in the board, school size, and number of disadvantaged students in the school). Furthermore, it explores nonlinearity in the second-sixth-grade relationship and examines the development of first-grade low and high achievers throughout primary school. 63
RkJQdWJsaXNoZXIy MjY0ODMw