c . They were listed in the summer 2019 partnership monitoring return to the OfS , as schools or colleges with which partnerships had begun delivering intended outreach activity .
d . They were recorded in the summer 2020 tracking report data , as schools or colleges that received at least 10 hours of engagement from Uni Connect partnerships .
e . They were listed in the summer 2021 tracking report data , as schools or colleges that received at least 10 hours of engagement from Uni Connect partnerships . f .
Over half of the learners at a given school or college across the six Key Stage 4 cohorts were from Uni Connect areas .
g . More than 100 learners at a given school or college across the six Key Stage 4 cohorts were from Uni Connect areas .
10 . Furthermore , in light of the finding in Annex A that gaps in application rates are only apparent at higher levels of prior attainment , learners with fewer than five grade A * to C ( or 9 to 4 ) were also excluded from this part of the analysis .
Sensitivity analysis
11 . We performed a range of checks and sensitivity analyses to ensure that the findings from the matched counterfactual analysis were robust to changes in our approach . We considered the following aspects :
a . We experimented with matching on fewer characteristics , for example excluding English and Maths results , or matching on GCSE attainment only .
b . We tried matching on both IMD and IDACI in turn as alternative measures of disadvantage to FSM .
c . We tried matching on school type ( within all state-funded mainstream schools or colleges in England ) in addition to the other criteria .
d . We experimented with relaxing the threshold for excluding learners from non-Uni Connect areas to minimise the spillover effects from potentially being in a Uni Connect targeted school or college . The condition that the school or college must have 100 or fewer Uni Connect learners over the four years was increased to 200 and 300 or fewer Uni Connect learners .
e . We concluded there was no need to test the exclusion threshold of 50 per cent of Uni Connect learners at a school or college as there were few schools or colleges with a very small number of learners over the years which were not otherwise excluded . f .
We ran a statistical model of the form adopted in the previous publication of this analysis . 46 Given the overall conclusions were the same as those from the matched counterfactual approach , it was decided not to report the results of that model in this
46
Available at www . officeforstudents . org . uk / publications / uni-connect-national-evaluation /.
64