That Theos Research on Social Class and Faith/Atheism

Since Church Mouse noted the latest Theos press release on social class and atheism, some other people have been picking up on it. Hold on a sec'.

The full research is here - it's actually a subset of their bigger piece of work on Faith and Darwin earlier this year. The figures on atheism and education are....

(first number is the count, percentage is the proportion of people with that qualification level)
"I believe in God but have not always done so"
Graduate degree or above 43 (10%)
Bachelors degree 14 (7%)
AS/A level 28 (9%)
GCSE or equiv. 30 (6%)
BTEC 2 (6%)
NVQ 9 (7%)
No quals 12 (5%)

"I used to believe in God but no longer do so"
graduate degree or above 34 (8%)
Bachelors degree 25 (12%)
AS/A level 21 (6%)
GCSE or equiv. 37 (8%)
BTEC 1 (3%)
NVQ 14 (11%)
No quals 25 (9%)

here was the Theos conclusion
converts to atheism are disproportionately drawn from groups with no educational qualifications whereas converts to theism are disproportionately drawn from higher socio-economic grades and from people with a masters degree or higher.

now I've got into trouble before on this blog for talking about stats as though I was an expert, and getting it wrong (ish). But I'm not really sure you can draw any firm conclusions from numbers this small. And if you take Bachelors degrees along with graduate degrees, there's actually no difference between the 2 questions at all for the 'most educated'.

Here are the findings on class:

I believe in God now but have not always done so
AB 44 (8%)
C1 49 (8%)
C2 28 (7%)
DE 36 (7%)

I used to believe in God but no longer do so
AB 42 (8%)
C1 49 (8%)
C2 32 (7%)
DE 49 (10%)

The only significant difference here is in the DE grouping, but I'd be reluctant to based headlines and detailed analysis on the results of 13 people. My feeling is that the press release over-eggs the findings, but I'm happy to be corrected by people who know what they're talking about. If your sample size is roughly 400, what's the margin of error?