Wellbeing in Aotearoa New Zealand: A Population Segmentation Analysis

This paper seeks to understand which factors are related to differences in subjective wellbeing within the New Zealand population aged 15 years and above. It uses data from the New Zealand General Social Survey (GSS) and regression tree analysis to identify the factors most strongly related to differences in subjective wellbeing (SWB). The tree analysis divides the population into groups or segments, whereby people in the same segment share a similar level of SWB and the factors most strongly related to SWB. The analysis shows how combinations of factors explain differences in SWB within the population and provides a person-centric view of wellbeing across multiple dimensions of wellbeing. We use self-reported life satisfaction (measured on a scale of 0-10) as a measure of subjective wellbeing and as the dependent variable in a regression tree analysis. We include the characteristics used to construct the wellbeing domains and sub-domains derived in earlier work on multi-dimensional wellbeing by McLeod (2018) based on the Treasury’s Living Standards Framework. We also include additional characteristics available in the GSS and from linked administrative data. Having identified the population segments, we describe their socio-demographic and other characteristics. Owing to some significant changes to the health questions collected between 2016 and 2018, we include separate results for GSS 2018 and GSS 2014-2016. Tree-based methods use a series of splitting rules to split the population into groups. The total population is first split into two groups (or branches) using the variable that most strongly differentiates subjective wellbeing in the population. Each branch is subsequently split into two, these are split again and so on. The tree-construction process continues until further splits do not explain significantly more variation in subjective wellbeing or when a minimum group (or leaf) size is reached. The final tree is a partition of the population into several groups or segments. A key finding from this analysis is that mental health, having enough income to meet everyday needs, and trust in institutions or trust in other people, are the characteristics most strongly related to different levels of subjective wellbeing in the population aged 15 year or above. While our results are somewhat sensitive to the survey year, owing to changes to the health questions collected, we find that of all the characteristics we consider, mental health consistently appears at level 1 in the trees and is the characteristic that is most strongly related to different levels of subjective wellbeing in the population.