Friday, August 2, 2013

Lindsay Mitchell: CPAG research inconclusive at best

New research from the Child Poverty Action Group, who are campaigning to increase benefit levels, finds there is,
 "no evidence of an association between benefit receipt and distinct substantiated rates of child abuse."   
And later under 'Conclusion', "...benefit income does not appear to be related to rates of child abuse."

How did they arrive at this?

Here's a cut and paste of the entire relevant text and scan of their scatterplot:

3.5 Benefit uptake

Benefit data is held by the Ministry of Social Development. We have assumed WINZ offices cover broadly the same areas as CYF site offices as MSD appears to know what percent of the population is on a benefit (see for example Collins, 2013) so would use the same or similar population estimates they use for CYF purposes. Here, data for the four main income-tested benefits has been combined since not all the jobless go on an unemployment benefit nor is it only DPB recipients who care for children (similarly, not all DPB recipients care for children, although most do). Benefit figures are from June 2012 so are for a point in time only: the numbers can easily change if the dynamics of people moving on and off benefits changes. The percentage of beneficiaries in the population was plotted against the rate of substantiated distinct abuse cases in each site office. If there was a relationship we would expect to see rates of substantiated abuse rise as the proportion of working-age beneficiaries rose. This relationship is statistically insignificant, with the R 2  value being 0.39. The values for each site office are in the Appendix at Table 6. 

Figure 5: Scatterplot showing rate of distinct substantiated cases of child abuse and proportion of income-tested beneficiaries in population for CYF site offices. R 2  = 0.39


The site offices with rates above 2.5 include Taupo (2.76 but a below-average proportion of income-tested beneficiaries in the population); Clendon (2.96 and a slightly higher than average proportion of income-tested beneficiaries in the population); and Whakatane (3.5 and an above average 10.1% of the population being on an income-tested benefit) and Papakura (4.0 with the 10% of the population estimated to be on an income-tested benefit being slightly less than Whakatane). There are 8 site offices with a higher proportion of the population receiving an income-tested benefit that have below average rates of substantiated child abuse.

The weak relationship between benefit receipt and child abuse may be no more than a reflection of the impact of the low incomes of benefit recipients (Perry, 2007). The data here shows no evidence of an association between benefit receipt and distinct substantiated rates of child abuse.

I have one major problem with this chart. It uses benefit dependency rates across all benefits. Most people on a working-age benefit are not caring for children (58% 2012). Most children dependent on a benefit rely on the DPB (78% 2012). What would the chart look like if the relevant population was plotted?

Earlier Auckland University research found,
 "There are 10,300 children maltreated by age 5 and seen on a benefit by age 5, comprising 86% of all children with maltreatment findings by that age – suggesting that the majority of children  with maltreatment findings are on the benefit relatively early in their lives since we only lose 3% when we ignore children who arrive on the benefit after age 2."
This strongly suggests that most would appear on the DPB. Consider this graph from the Children's Social Health Monitor:




Anyway, CPAG took exception to the Auckland University in their first report saying,
"It suggests child maltreatment is a function of membership of particular social groups, something for which the evidence is very weak."
They footnote this claim with the following, "See Child Poverty Action Group technical paper (forthcoming)."

And here it is.

Whose research is more robust? The University of Auckland or the Child Poverty Action Group's?

No comments: