Note: The graph shows the DfES KS1-KS2 value-added scores, and the average
points score per pupil at Key Stage 2 for all primary schools in York, Leeds, and
North Yorkshire LEAs.
Both the KS1 and KS2 scores, of course, contain a considerable but unknown element
of error. The Key Stage tests may have less than perfect validity in what they
purportedly measure, candidates may make untypical mistakes in responding to
questions, some teachers and schools may condone ‘sharp practice’ in administering
the tests, some candidates will be missing, and some candidates will have missing
scores. There may be mistakes in the marking, recording and computing of the KS2
points per school. The marking is to a threshold in which the achievement of two
pupils just above and below a threshold may actually be closer than the achievement
of two pupils awarded the same grade. The grades are converted to a points score,
which changes the metric and may create additional distortions in the data. The value-
added scores are then created from these two imperfect sets of figures, and the value-
added model is only one of many possible, requiring a number of untestable analytical
assumptions based on subjective judgements. This level of uncertainty in the result
could be sufficient to explain the apparent differences between the value-added scores
of schools with similar raw-scores in Figure 1.
The correlation between the primary schools’ value-added score and their KS2 results
is +0.74 (Pearson’s R). One of the major reasons why this correlation is lower than
that previously published for secondary schools (Gorard 2006c) is that primary
schools are generally much smaller, with fewer pupils in each cohort. Therefore, there
is more volatility in the figures (or put another way, the measurement problems
outlined above are more apparent - see also Tymms and Dean 2004). One way of
assessing whether this is the correct interpretation is to examine the correlation for
large and small primary schools separately. If the correlation is lower for small
schools but larger for large schools then this is an indication that the volatility of
small schools helps makes the correlation ‘appear’ smaller than it is at the secondary
level.