Examnishambles rumbles on
David Kernohan is Deputy Editor of Wonkhe
Tags
A great story from Schools Week saw a little bit more data and information leak out from DfE about the examnishambles of 2020. We get confirmation that DfE knew about the widespread algorithmic downgrading before the first set of results were released, and that even though the “calculated grades” were on a part with 2017, their main concern at the time was to identify schools who awarded all their students A*s.
We also get to see the extent of “calculated” downgrading (and, in a few rare cases, upgrading) by A level subject – so I couldn’t resist a few graph. This first dashboard shows the information on the proportion of adjustments made from the centre assessed grades to the “calculated grades”, basically what the PM called a “mutant algorithm” wrought.
And here I rank subjects by the proportion of downgrades, with the number of candidates affected in the row below (grey here is candidates not affected, and green is upgraded, by the algorithm).
We’re hardly short of data as to how alarmed the general public were that A level grades (in every year, not that this particular aspect was the one the cut through!) have a lot more to do with a pre-determined curve of results adjusted for demographics and regional differences, and a bit less to do with actual candidate performance than is generally understood.
Last year the teachers I’ve spoken to (not a representative sample, I hasten to add) mentioned that they used the grades awarded to their previous cohort as a guide to understanding what grades they would expect this years’ entrants. This was only one factor in determining a grade, and an eminently sensible one, but it also illustrates why grading to a curve is not a great idea. The standard of work needed to get an A in 2016 wasn’t the same as the standard needed in 2017 – two entirely normal years. Why would it be the same in 2020?
These charts show you in which subject areas that teacher expectations (and thus, arguably, candidate performance) were out of line with previous years. The large number of algorithmic downgrades suggest that this crop of A level students were working at a very high historic level.