The descriptive detail of the last chapter laid out how repeaters' test scores and test score gains compare with promoted children's, as viewed from several vantage points. Such comparisons, however, are a step removed from telling us whether grade retention abets or inhibits children's academic progress because repeaters and promoted children differ in a host of other ways besides their retention status – for example, low family income and weak preretention academic skills. In an evaluation, these factors constitute possible confounds – they, rather than the retention experience, may lead to observed outcome differences. Until they have been addressed we cannot draw firm conclusions about retention's impact.

The strategy of matched controls can be used to isolate effects of retention from other considerations; however, as discussed in Chapter 2, not all alternative explanations can be disposed of under this approach. Our repeater and comparison groups are (roughly) comparable in terms of early testing levels, but retainees and promoted children differ in many other ways not equated by the California Achievement Test (CAT) match. An alternative to creating matched groups is to adjust statistically for characteristics deemed relevant. The statistical approach has the advantage of being able to adjust for more factors than is practical by matching and poses less of a threat to accurate significance levels.

Retainees are disproportionately members of low-income families; they have poorly educated parents; they are often of minority status; and they enter school with weak readiness skills, reflected in low beginning test scores and marks.