Another doubtful league table?
David Cameron has prominently commented on the recent performance tables concerning adoption in local authorities, in particular the proportion of children whose adoption placement occurs within 12 months. But are the local authorities really as different as they have been made out to be?
The league tables are available here but only give percentages. Considerable searching is necessary to find the raw numbers on which these percentages are based, but it is possible to eventually discover details here under ‘local authority indicators at LA level’ Table I2. This spreadsheet only gives numbers rounded to the nearest 5, although the percentages given in the league table use the unrounded numbers. By adding the rounded denominators (and substituting ‘5’ for denominators which are not given at all), a spreadsheet can be constructed for data for 2008-2010 in 143 authorities – this is available from Googledocs.
Just as in an analysis of bowel cancer mortality rates , which was also covered in a recent article by Ben Goldacre, we can use a funnel plot to display the data. The one below has been constructed simply by cutting and pasting the final 3 columns of the spreadsheet into ERPHO’s excellent funnel plot generator.
Most of the local authorities (LAs) lie within the 95% (2SD) funnel and so their variation is essentially indistinguishable from chance. Many of the more extreme authorities handle low numbers (eg York was top of the table but only handles around 10 adoption a year): many of those at the top of the league table are similarly small and have performance that is not significantly different from average.
While York and 5 other Las can be considered as having systematically better performance, only a few LAs can be confidently said to be worse than average.
Although there is clearly room for improvement, it seems inappropriate to construct a league table based on such low numbers. In addition, highlighting this one performance measure obscures the overall picture obtained from the multiple indicators.