Success rate, sample and rating distribution
Out of a total of 171 evaluations, 129 were included in the success rate calculation (only projects from the drawn sample), 63 of which originate from the 2019/2020 sample. The remaining 66 projects originate from the samples from previous years.
The majority of the 63 projects from the 2019/2020 sample were rated of 2 and 3, though the percentage of projects rated as 2 was slightly higher (51% rating 2, 38% rating 3). The remaining projects were rated as 4 and 5 to roughly the same extent (5 % and 6 %, respectivley). No ratings of 1 were awarded. Projects are deemed successful if they receive a rating of 3 or above.
The distribution of ratings is therefore similar to the distribution in previous periods but with some small changes: while there are significantly fewer projects with a rating of 4, the percentage of projects rated as 2, 3 and 5 has increased. The top rating of 1 remains an exception.
Looking at the distribution of ratings by volume of funds gives a similar, but more volatile picture over time. One striking feature of the 2019/2020 sample is that, taking into account the volume of funds, a larger proportion was rated 3.
Accuracy of estimates and confidence interval
The success rate for the period of 2019/2020 is slightly higher than in previous years but does not exhibit statistically significant differences.
On the whole, the success rate has been trending upwards since 2011/2012. However, this is not statistically significant, nor is the slightly lower value for the estimated success rate in the 2017/2018 period.
The percentage of completed evaluations from the 2019/2020 random sample is still comparatively low. As such, the estimated success rate for the universe of evaluations from 2019/2020 has yet to be finalised.
By evaluating remaining projects from earlier samples, it was possible to increase the accuracy of estimated success rates for previous periods. For example, 56 evaluations from the random sample for the basic population for 2017/2018 were completed in the 2019/2020 period, which meant that the range of the confidence interval could be reduced by around eight percentage points.
