Saturday, October 13, 2018

Breaking: MSD won't publish latest child poverty rate due to "uncertainty" and "lack of confidence" in data

Big news not reported in mainstream yet.

From the latest Household Incomes Report, (Headline findings), the source of official statistics for child poverty:

The 2018 reports do not publish low-income and material hardship rates for children for 2016 and 2017:
o Last year’s reports noted that several of the key rates for children for 2016 were surprisingly low compared with the relatively flat stable trend for the previous three years and warned against reaching any definitive conclusions on the short-run trends using the 2016 figures. The 2017 figures are much the same as the 2016 figures. There are no known factors in the economy, the housing market or policy change that can explain the falls to 2016 and 2017. While sampling error can account for some of the difference, considerable uncertainty remains.
o Stats NZ is scheduled to report on these statistics for children in their new Child Poverty Report in early 2019, using more up to date survey information, supplemented with administrative data.
o MSD has therefore decided to take a pause on reporting these rates for children in the 2018 reports. Stats NZ supports this cautious approach. 
You can read the Minister's briefing for detailed explanation which essentially blames sampling and non-sampling error. This sums it up:

"...the 2016 and 2017 samples may have some sample bias away
from poorer households with children. As noted above (para 17), one way that sample bias can occur is through non-responders being different from the responders in important ways that are not addressed by standard weighting procedures. If, for example, it proves more difficult to get responses from households with low incomes or high material hardship than it does to get responses from better off households, then the sample is likely to be biased and the bottom end will likely look better off than expected. The investigation to date is not conclusive on this, and does not explain why it suddenly appeared, but it does point to something unusual happening with the samples."
To not publish their data is quite extraordinary though labelled a temporary measure:

44 We have briefed the following parties on the decision to not publish low-income and
hardship rates for children, and the rationale for that decision: your office, DPMC, the Child Poverty Unit, the Prime Minister’s Office, the Minister of Finance’s Office and the
Treasury.
45 The relevant staff at MSD and Stats NZ have also been briefed.

Looks like the media has missed this. As per usual the report was put on the MSD site Friday.

Here's the big deal about this.

If the data can't be relied upon for whatever reason what is the point of the Prime Minister's Child Poverty Reduction Bill? Her self-proclaimed raison d'etre.

Sepuloni says:

This year’s report does not include low-income and material hardship figures for children in 2016 and 2017 because of sizeable changes in levels that officials cannot fully explain, even when the relatively small sample size (3500) is taken into account. More information can be found here
From next year, the Incomes Report will use improved data from Stats NZ. From 2020, it will provide greater precision by drawing on a sample of 20,000 households.

You could draw on 100,000 but if the non-responders are disproportionately poor the results will be skewed. 

4 comments:

Kiwiwit said...

This is like something out of the Soviet Union, where the authorities used to stop publishing the 5 year plan figures whenever they went the wrong way. The child poverty measures, being relative (which means they never improve with improvements in overall living standards) are designed to justify socialist ends of equality over all else, but even then the Government decides not to publish them when the trends go the wrong way. Obviously the previous government was doing something right.

Murray Roxburgh said...

Even more embarrassing will be the effects of the falling dollar and the avarice in the fuel taxes and virtue signalling that will arise before the ink is dry.
The lower socio economic sector are going to be hit hard.

Jim Rose said...

Report was several months late too

Lindsay Mitchell said...

Yes. MSD says "The reports usually come out in July, but the timing is very dependent on when we receive the Household Economic Survey (HES) data. This year the data was very late and the relevant resources were also working on the Child Poverty Reduction (CPR)
Bill."

How are they going to fix that when they want to do even more surveying to get 'better' data? Increasingly I think surveying is a nonsense. In the case of the HES they get around 80% response. It's not good enough information to draw sound conclusions even if it is an internationally good result. So what?