Friday, August 30, 2013

Lindsay Mitchell: Child abuse rates in the beneficiary population: MSD cover-up by omission

Warning: laborious statistical workings below...

I've hesitated to label a new Ministry of Social Development factsheet a "cover-up" but having chewed over it for a few days I've decided that's exactly what it is. For the first time MSD has examined a "birth cohort ever present in New Zealand" and the overlapping contact with the benefit system, care and protection or youth justice services in the years to age 17. They used 1993.

The opening summary states:
"While most supported by the benefit system in childhood had no contact with care and protection services, as a population group their likelihood of contact was 1.5 times that for the cohort overall."
If you think that seems low there's a reason. Throughout the factsheet the benefit group is compared to  the total cohort (which includes the benefit group). Why not compare the benefit group to the non-benefit group?

Maybe because the result doesn't look so good.

Let me demonstrate:

 Here's MSD calculations:

Of 41,000 benefit cohort 12,000 (80% of all known) had contact with C&P =.293
Of 76,000 total cohort 15,000 (100% of all known) had contact with C&P = .197

.293 is 1.5 times .197

But let's look at the non-benefit cohort

Of 35,000 non-benefit cohort 3,000 (20% of all known) had contact with C&P = .086

.293 is 3.4 times .086

Children who appeared in the benefit system were 3.4 times more likely to have contact with C&P than non-beneficiary children. Nowhere in the factsheet does this likelihood appear.

Being known to C&P doesn't necessarily mean a child has been abused or neglected. There is however also data available on the incidence of substantiated findings.

The paper says "looking at the estimated 18 percent of the cohort who spent at least nine years of their first 17 years supported by the benefit system...27 percent had substantiated findings of abuse or neglect (a rate more than three times that for the overall cohort...)"

Again the authors make the comparison to the overall cohort and not the non-benefit cohort. Let's do that with the data available in the next table:

We know from the previous table that 27 percent of the 9+ years benefit group had at least one substantiated finding of  abuse or neglect. The 9+ group represents 18 percent (table above) of the total cohort of 76,000 or 13,680.  So 3,694 of them had a finding.

3,694 is 60 percent of the total children with a finding. Total children with a finding must therefore be 6,157.

3,694 findings (60%) amongst 13,680 (9+ benefit group) = .27
739 findings (12%) amongst 35,000 (non-benefit group) = .021

.27 is 12.9 times .021

Children who spend 9 or more years in the benefit system are almost 13 times more likely to experience a substantiated finding of abuse or neglect than children never on a benefit. That's more meaningful than a likelihood, "more than 3 times that for the overall cohort."

Yet turning to the "discussion" section of the factsheet, MSD says:
"The findings are consistent with associations between low income and measures of child maltreatment found both across and within countries. They do not however establish that being supported by the benefit system causes a child to be more at risk of these outcomes."

Right out of the CPAG handbook. It's a pretty pointless exercise then, analysing benefit and CYF overlap only to write the association off to "low income".


Anonymous said...

Long time reader, keep up the great work. You seem to be a lone voice in the wilderness actually challenging widely accepted philosophies and assumptions. You have a critical voice - please keep doing this.

Anonymous said...

Hi Lindsay - I concur with Anon above.

Stuart L

Anonymous said...

Would someone please explain to me how making an issue appear better than it really is help resolve the issue. Surely it is in everyone's best interests to be honest about how bad the problem is and then work at solving it.

Ken said...

Well written as always Lindsay.

Unknown said...

What about challenging Paula Bennett on your findings or better still John Campbell on TV3. Leaving the at risk group in the non-risk group is either a deliberate skew or ignorance of 6th form stats.