we show that less than one in five learners who were living in target areas in the most recent cohort were recorded as having attended a school which later received the full four years of engagement available to them . On the other hand , nearly all these learners are understood to have attended a school which received at least some engagement .
e . This data does not identify individuals who have engaged with other outreach programmes . Therefore , there will be some people who have benefited from other outreach programmes who we then compare against learners from Uni Connect areas . f .
Similarly , we cannot take into account any pre-existing outreach targeting the same outcomes and areas as Uni Connect , which either stopped when Uni Connect was launched or became incorporated into the Uni Connect programme . This means that the year used as a comparison in this analysis ( 2016 – one year before the launch of Uni Connect ) relates to a cohort which may have benefited from other outreach activity , while more recent cohorts did not .
g . To analyse cohorts earlier than would otherwise be available through the administrative higher education entry data , we based this analysis on applications through UCAS at age 18 for full-time undergraduate courses . But this is more narrowly defined than higher education participation , which includes more courses ( such as part-time courses ) and entry by age 19 . Annex C begins to address this gap in evidence by using the available higher education data .
h . The quantitative administrative data does not capture all factors that are associated with participation in higher education . Therefore , even after matching learners on a set of personal characteristics as we do in the final section of this report , there will always be other factors that are also associated with higher education participation which remain unbalanced across the two groups . For example , the data does not capture the level of school or parental support each learner received , or differences in attitudes of individual learners . As previously stated , the data cannot capture all differences in pandemicrelated behaviour between these two groups either . i .
Part of this analysis matches learners from Uni Connect areas with those from other areas according to the number of GCSEs they held at grades A * to C ( or 9 to 4 ). Although it was not an aim of the Uni Connect programme to raise attainment among participating learners in either phase one or two , it is possible that it has indirectly had this effect for the two most recent cohorts of learners , whose Uni Connect engagement potentially began in school years 9 and 10 respectively . It is therefore possible that matching learners on GCSE attainment disguises any knock-on effect of higher attainment on application outcomes .
39 . Overall , these limitations mean that the analysis in this report cannot identify a truly causal relationship between the Uni Connect programme and trends in higher education participation at a national level . It does identify whether national gaps in participation between the most and least represented areas are closing , whether this is evident in areas targeted by Uni Connect , and whether these trends appear to be associated with other underlying differences in characteristics between groups of learners . But it cannot attribute causality to the Uni Connect programme .
16