What we don't know is probably more relevant to policy

 A just-released MSD paper examines the family structure of Maori children.

Two data sources were used, Census and Growing Up in NZ. 

First the Census data:

You will notice that these numbers cannot represent all children. They are too low. The paper explains:

We selected children who were able to be linked across all three censuses by Stats NZ. Overall, 59 percent of the child population aged 0–4 years in 2001 and captured in the linked census files was linked to data in the two subsequent censuses (2006 and 2013). Among tamariki Māori, this drops to 53 percent. Some of the ‘missed’ links can be attributed to demographic factors, such as emigration and mortality, but a larger proportion of these false negatives are likely due to incomplete or inconsistent identifying information on children, which means they are not able to be linked. 

The paper states:

"... the share of tamariki Māori ever living in a family with two-parents only (67.5 percent) was significantly lower than among all Aotearoa NZ children (80.2 percent)."

I am wondering why a comparison wasn't drawn between Maori and non-Maori? So I made another table subtracting the Tamariki Maori data from the All Children data.

Now the gap between Tamariki Māori ever living in a family with two-parents only (67.5 percent) and non-Maori NZ children (84 percent) widens.

Another finding (mine) 50.1% of Maori children aged 0-4 in 2001 were in a non- two parent family. This compares to 30.6% for non-Maori.

But this is still tenuous stuff due to all the missing data.

And it gets worse with the GUiNZ data.

It's probably enough to quote from the paper:

These [GUiNZ] longitudinal data allowed for the examination of family structure over multiple time points across early childhood. In this report, we examine family structure data available at antenatal, and when the focal child was 9-months, 23-months (ie nearly 2-years old) and 45-months old (ie nearly 3.5-years old). Family structure was not available at the 54-month wave (ie when the child was nearly 4.5-years old) – the wave in which child outcomes were measured. In total, we were able to include family structure measures at four time points.

In the externally available GUiNZ dataset, family structure is coded by the GUiNZ research team into four mutually exclusive groups from a household roster reported by the primary respondent (mostly the biological mother). These are:

1. living with two parents and no other adults

2. living with one parent and no other adults

3. living with one or two parents, and other adults who are kin

4. living with one or two parents, and other adults who are not kin (and potentially other

adults who are kin).

It is important to note four primary limitations in this conceptualisation of family structure. (My emphasis)

I am constantly frustrated by data limitations because relationship status between parents, and parents and children in some cases, is ignored.

I am not a political conservative. But science finds committed parents (mostly manifested through a marriage) stay together more than any other co-producers of children. Their children demonstrably benefit from this. Most sociologists - and governments by extension -  are impervious though.