Grade retention, socioeconomic determinants, and research quality

Abstract

The relationship between repetition and socioeconomic and cultural level has been widely studied in Spain; however, the estimates offered are varied, even conflicting. One reason for these divergences may be the difficulty in neutralizing two types of biases inherent in repetition studies: selection bias and covariate control. The aim of this research was to estimate the probability of repeating in three social status groups: low, middle and high. A sample of 5999 students was used, assessed at two times (T1 = 4^th grade; and T2 = 8^th grade). In T1 the sample was from the same grade and age: the students had not previously repeated and the first repetitions did not occur until the end of 6^th grade. To control for selection bias, the groups (repeaters and those promoted) were matched in T1 using a propensity analysis and in T2 the matched estimates were combined with the adjustment of the data to multilevel logistic models. Specifically, the predictions of three models were compared, each defined by the degree of control included in its design: (1) descriptive, without control variables or equivalent groups; (2) transversal, only with covariate control at T2; and (3) longitudinal, with double control, meaning equivalent groups before repetition plus adjustment variables at T2. The results showed that the model with better bias control (longitudinal) presents more moderate and partially significant effects. The data highlight the need to develop, in our country, specific research on grade repetition that includes repeated measures before and after repetition with matched samples, or the alternative use of quasi-experimental approaches. In this way, the debate on grade repetition will be enriched with the contribution of fully valid data.

Key words: repetition, education, socio-economic status, hierarchical models, propensity score.