Simply observing that a sixth-former is more likely to choose a particular course when a former pupil of the same school was admitted to it the previous year is not sufficient
to establish a causal link between successive cohorts. It is, in fact, to be expected that pupils from the same secondary school will frequently apply for the same courses, if only because of their geographical proximity. Furthermore, the counterfactual scenario – the choices pupils would have made had alumni been allocated to other courses – is never observed.
To measure the causal impact of former pupils’ admissions on the course choices of subsequent cohorts, the approach proposed here involves comparing, for a given year, secondary schools where a pupil was admitted to a particular course (so-called ‘beneficiary’ schools), with very similar secondary schools where no pupil was admitted to the programme in question, even though there were applicants (so-called ‘control’ schools). To define these two groups of secondary schools, the authors use the method known as ‘discontinuity regression ’ method. For each course, the rank of the highest-ranked applicant in each secondary school is determined, and the secondary schools are ranked according to this rank. The school at which this candidate is the last to be admitted to the course is assigned a rank of zero. Secondary schools whose top-ranked pupil is admitted are assigned a positive rank in the defined order (+1, +2, etc.),
and those whose top student is not admitted are assigned a negative rank (−1, −2, etc.), as illustrated in Figure 1.
This approach is similar to that used by Estrada et al. (2025), who analyse the impact of former pupils on applications from younger pupils to elite secondary schools in Peru.
The analysis then focuses on secondary schools where the top applicant is in the immediate vicinity of the admission threshold (within 20 ranks of this threshold). As these thresholds cannot be anticipated by either applicants or schools, the fact that a school’s top-ranked pupil is just above or just below the threshold is virtually random. Secondary schools situated around this threshold therefore, by definition, share almost identical characteristics – an assumption validated by the data.
Secondary schools situated just to the left of the threshold, whose top-performing pupil was narrowly rejected, constitute the control group, whilst those situated just to the right, whose top-performing pupil was narrowly admitted, form the beneficiary group. As some offers of admission may be turned down by applicants, only a subset of secondary schools see a former pupil ultimately admitted to the relevant course: this is the case for around 22 per cent of the beneficiary secondary schools.
