OBSERVATIONS ON HOME EDUCATION STATISTICS
TAKEN FROM THE HOUSE OF COMMONS CHILDREN, SCHOOLS AND FAMILIES
Meeting started on Monday 12 October at
Prepared by Jonathan Camfield
MA (Cantab) (Mathematics) and Fellow of the
Institute of Actuaries
Home educating father (believed to be one
of those that the local authority do not know about)
This note explains why Graham Badman's statistical
conclusions on the incidence of "at risk" children in
the elective home education (EHE) communityapparently broadly
twice the incidence of that in the wider communityappear
to be flawed. This note has been prepared solely by reference
to the discussion held at the Select Committee hearing on 12 October
At 1:08 in the television recording of
the Select Committee, Graham Stuart MP asks some questions about
the statistics for "at risk" children in the EHE community.
What is clear from the earlier questions and
answers is that extensive work has been done on this by Graham
Badman's team. In particular:
74 local authorities (around half
of all local authorities) have provided detailed figures (providing
a helpfully large database); and
It is clear that children at risk are
being defined carefully and quite narrowly in the statistics,
ie only those who are subject to a child protection plan (CPP)
under Section 47 (that is, it does not include children who
are simply disabled; they must have been formally assessed as
"at risk" by a local authority and reached the end of
that assessment process).
On the basis of the statistics, Graham Badman
has identified that 0.2% of the whole UK child population are
subject to a CPP whilst 0.4% of EHE children are subject to a
CPP. Whilst it is accepted that there is a wide range of statistics
from different authorities, Graham Badman has reached the broad
conclusion that the risk of a child being "at risk"
within the EHE community is double that in the wider population.
However, as pointed out by Mr Stuart shortly
thereafter in the discussion, this analysis appears to be statistically
flawed. In particular, there are three known figures, and one
unknown figure. The following are known:
The number of children in the whole country
is broadly known;
The number of CPPs in the whole country
is broadly known;
The number of CPPs from EHE families
is broadly known (in that, by definition, a CPP must be known
to a local authority and that local authority will also know whether
or not that child is at school).
However, the number of home educating families
is not known. Graham Badman states that the estimate for known
EHE children is around 20,000, but that most local authorities
believe that there are at least double that number. If there were
in fact double, then the crude statistics for the incidence of
CPPs amongst the EHE community would revert to 0.2%. If there
were in fact (as is believed by local authorities) at least double,
then the statistic would fall to less than that of the general
After misunderstanding each other for a few
minutes, Graham Badman understands this point as it is made by
Graham Stuart (at 1:13:20). However, Graham Badman goes on to
make a seemingly controversial statement. He says "Who is
to say that they [the children] are safe if you don't know anything
about them?". The point he is trying to make is that, in
the "unknown" EHE community, there may be more children
who, if they were known about, would indeed swiftly find themselves
subject to a CPP. (He makes this point again very clearly at 1:34 when
the Committee re-sits after the break.)
On the basis of the information available to
me, it appears to me that Graham Badman's logic is statistically
flawed. In particular, he is making an assumption that:
Both "lack of knowledge" to
a local authority and being a EHE family;
Puts a child more at risk (indeed, doubly
Both "lack of knowledge" to
a local authority and being a non-EHE family (ie child at school).
This appears to be a clear bias in his view.
In particular, it seems clear to me that out
of all the unknown EHE families, there will be children at risk
who should in fact be subject to a CPP.
However, it also seems clear to me that out
of all the non-EHE families whose children go to regular schools,
there are also a number of children at risk who the local authority
does not know about, and who should in fact be subject to a CPP.
Whilst Graham Badman is clearly entitled to
his personal views it seems unreasonable to me for him to suggest
that the statistics demonstrate something for which there is no
statistical evidence. That is, to suggest that the incidence of
being "at risk" is greater amongst "unknown"
home educating families than it is amongst "unknown"
regular schooling families. This is even more the case when the
crude statistics themselves (after doubling the numbers of estimated
home educated children) suggest that there is in fact no difference
in incidence between the two groups.
This conclusion (that there is no higher risk)
is stated clearly by Graham Stuart at 1:13:40. I believe that
his statement at this point is 100% correct based on the raw statistics,
and that Graham Badman has drawn a statistically incorrect conclusion
from his data. He has omitted the fact that the number of home
educated children is estimated to be at least double the number
actually registered and, at the same time, he is coming to a strongly
biased conclusion about unknown children.
After the break Graham Badman states that he
will write to the Committee to confirm his statistics, including
his view (which I note is a personal view, not a statistical one)
that his statistics are correct because not being known to a local
authority and being an EHE family puts you MORE at risk. It seems
clear to me that it is unreasonable to draw this conclusion from
the data he has available to him.
As a final aside, what the whole discussion
on statistics also does not take into account is the statistical
feature of "cause and effect". That is, there may be
good reasons why a child with a CPP is more likely to be either
at school or home educated. That is, even if the incidence of
children with CPPs is in fact higher in one community compared
with the other (eg in the EHE community), this does not necessarily
mean that in general home educating families are in some way more
at risk of having a child subject to a CPP (ie it is unreasonable
to assume that home education is a causal factor in children being
"at risk"). What is far more likely to be true in such
a case is that there are other reasons (alcohol, drugs, learning
disabilities in the home etc) which both:
(a) make it more likely for a child to be subject
to a CPP; and
(b) make it more likely for a child to be removed
from school (or at school, depending on what the statistics show).
However, given that the statistics appear to
demonstrate that there is no difference in incidence of CPPs amongst
home educators and the wider population at the crude level, this
does not seem to be a point worth exploring further. Normally
it would be a crucial point.
In conclusion, it appears to me (and I believe
Graham Stuart is of the same view) that Graham Badman's own statistics
clearly demonstrate that the incidence of CPPs amongst both home
educators and non-home educators are broadly similar, at around
0.2%, and that therefore government policy should not be formed
on the basis of there being a greater incidence of children at
risk in the EHE population. Indeed, this sad statistic might suggest
that it would be far better to allocate increasingly scarce resources
to children who are in fact at risk, rather than to the home educating
community or to the schooling community.