are equivalent to confidence intervals at the 95 per cent level ; we can be 95 per cent confident that the true change in the gap lies within this range . 38
116 . As Figure 14 shows , in the 2021 application cycle , there is statistically significant evidence that the gap in application rates has widened since the Uni Connect programme began in 2016 , by between 0.5 and 1.7 percentage points , even after accounting for differences in characteristics between learners from Uni Connect areas and other areas in matched characteristics . This contrasts with the movement in the previous application cycle in 2020 , where we are unable to conclude that the gap in application rates had changed at all since 2016 , with estimates ranging between -1.0 and 0.2 percentage points at the 95 per cent confidence level .
Figure 14 : Estimated percentage point change in gap in application rates since 2016 after taking into account differences in matched characteristics between learners
117 . It is worth recalling the context around these estimated changes . First , even the most recent cohort applying in 2021 had not yet had the opportunity to benefit from the five full years
38
Equivalently , if we were to take new random samples , we would expect the change in the gap to lie within that range 19 times out of 20 . We have made no adjustment for multiple comparisons in the calculation of these non-parametric intervals .
42