Understanding NZ's low-income households 🍋

Insights from StatNZ's household economic survey

We all have intuitive ideas of what being on a low income means, and the impact that it has on the financial difficulties people experience. These intuitions inform our reactions to policy proposals.

But it’s good to check intuition against the data. The Infrastructure Commission (Te Waihanga) has just published How much do we pay for infrastructure? Household expenditure on infrastructure services, which reveals some insights into NZ households.¹

The report looked at household income and expenditure, using the expenditure sub-sample of StatsNZ’s household economic survey (HES).² StatsNZ’s disclaimer applies:

These results are not official statistics. They have been created for research purposes from the Integrated Data Infrastructure (IDI), which is carefully managed by Stats New Zealand. For more information about the IDI please visit https://www.stats.govt.nz/integrated-data/.

Splitting households into disposable income quintiles

The researchers classified households by disposable (after tax) income.³ They ranked households by disposable income, then split them into 5 equal-sized groups (quintiles). A natural interpretation is that quintile 1 represents ‘the poor’, whereas quintile 5 represents ‘the rich’. As you’d expect, the proportion of income that a household spends on infrastructure services reduces as income increases:

Another reasonable prior is that richer households would spend much more (in dollar terms) on infrastructure services than poorer ones. The data bore this out, but the relationship was weaker than I might have expected.

However, to me the most interesting finding was that the variation in spending within quintiles was greater than the variation between quintiles. This post explores what might be driving this variability in infrastructure spending between households of similar incomes.

Interpreting disposable income quintiles

Equating disposable-income quintiles with relative richness or poorness of households has a few wrinkles. For example:

1. Households with similar incomes may nevertheless have very different income sources, reflecting, among other things, the life stage of their residents. Two households with the same income today might have very different incomes in say ten years’ time.

2. Some households have substantial imputed income (in the form of avoided current expenditure) that is missed by surveys, such as the value that arises from asset use (e.g. imputed rental income from living in a house they own).

3. Households tend to save during periods of higher income, building up financial or physical assets, whereas during periods of lower income they “dis-save”, drawing down (or running down) such assets. Some households may be “asset rich but cash poor”, with low disposable incomes but a high standard of living. Conversely, households dedicating a large proportion of income to building an asset base (e.g. paying off a mortgage) may have relatively high incomes but still experience some hardship in the interim.

The report explored these wrinkles further.

Categorising households by income class

The researchers split households into five income classes, based on the primary source of household income.

The next table shows households divided between the income classes and disposable-income quintiles. The row mean shows the percentage of households in the sample in each income class. Employed households dominate (67%), then Retired households (21%), followed by Welfare-dependent households (7%). Student (2%) and Other (3%) households account for the remainder.⁴

Retired, Welfare-dependent and Student households are strongly skewed towards the low-income quintiles. Employed households are skewed towards the middle and high-income quintiles.

Which households save? Or run down their savings?

Household income can be volatile from year-to-year. Individual households, to varying degrees, can manage this volatility by saving (e.g. financial savings, accruing & improving physical assets, investing in human capital, repaying debt) and dis-saving (e.g. drawing down financial savings, selling physical assets, deferring maintenance, increasing debt). Can we see evidence of saving or dis-saving in the HES data? To what extent is current household income sufficient to cover current household expenses?

To investigate, the researchers created a measure of “non-capital” household expenditure that excluded large financial and capital transactions (e.g. depositing into KiwiSaver, buying & selling houses or vehicles). Dividing disposable income by this measure creates an “expenditure coverage” ratio. A ratio below 100% means a household is dis-saving, whereas a ratio above 100% means a household is saving.

The next table shows mean expenditure coverage, by household income class and by disposable-income quintile. Cells ≤ 95% are shaded red. These households are, on average, drawing down on their savings. This is predictable and not necessarily a problem for Retired and Student households. But it is of more concern for Welfare-dependent households, and likewise for low-income Employed households.⁵

Cells ≥ 105% are shaded green; these households are, on average, increasing their savings.⁶

Do households have difficulty paying bills?

Recent HES surveys ask respondents whether they have had difficulty paying bills in the past 12 months.⁷ The next table breaks down responses by household income class and by disposable-income quintile.

The responses to this question appear inconsistent with the results reported above. For the population as a whole, and Employed households in particular, difficulty paying bills does not diminish quickly with income.

My own preliminary hypothesis, which would need to be tested with better data, is that payment difficulty is more dependent on budgeting skills than income; and it only reduces substantially once these bills represent a small proportion of total household income.

Another apparent anomaly is that reported difficulty paying bills rises with income quintile for Welfare-dependent households. Perhaps the combination of welfare-dependence and higher-incomes may reflect large numbers of household members, with relatively low per-person incomes and higher household budget coordination costs?

Difficulty paying bills is consistently low for Retired households, suggesting retirees are better at budgeting, place a higher priority on bill payment, or are under less financial pressure than their income levels would suggest.

Understanding household assets

The HES expenditure sub-sample has little data on assets. It does, however, identify the households that are renting vs. those who own the dwelling they live in. Overall, 68% of households were owners, ranging from 84% for Retired households, down to 18% for Student households. Home-ownership rates are more than 50% across all income quintiles, showing income by itself is a poor proxy for low household assets.

My further thoughts

The starting point for this analysis was the finding that variation in infrastructure spending within household-income quintiles was greater than the variation between quintiles. So clearly there was something else important going on. This analysis offers a partial answer, along with some interesting observations about the economic situation of New Zealand households.

First, disposable household income is an incomplete measure of financial pressure, financial distress, and low assets. It could be misleading if used alone for policy design. The lowest-income quintile is dominated by Retired and Welfare-dependent households, yet these household types differ dramatically on home-ownership and difficulty-paying-bills measures.

Second, financial distress affects middle-income households almost as much as it does lower-income households. This won’t be news to social-services organisations, who have long known the importance (and effectiveness) of helping people set and stick to a household budget.⁸

Third, households may run substantial deficits (running down assets) or surpluses (building them up) year-to-year. In general, this is a good thing — it reflects sensible planning and adjustment to circumstances. It is likely be problematic for those households with few assets to run down, or no means of rebuilding them to meet future need.

Last, it reminds us just how difficult it is to accurately target income-support policies. Targeting requires specifying criteria that divide households and people into eligible and non-eligible groups. Solid data and robust analysis is necessary for choosing criteria that achieve, rather than undermine, policy intent.

By Dave Heatley

Want more on this topic? Check out:

• Trends in the household income distribution: 2007-2021 (2023) by Meghan Stephens

• The material wellbeing of New Zealand households: trends and relativities using non-income measures (2021) by Brian Perry

• Expenditure patterns of New Zealand retiree households (2023) by Trinh Le and Euan Richardson

Frosty sunrise, Tararua Forest Park. Dave Heatley

1. This post is drawn from Appendix A Measuring household incomes. See the report for further results and methodology. I contributed to that work, under contract to the Infrastructure Commission. The interpretation and comments in this post are my own where they extend or vary from those in the report.

2. An enduring challenge for researchers looking at questions of financial distress is choosing the best unit of analysis. Income and tax can often be best understood at the level of individuals, whereas expenditure is more often considered an attribute of a household across which income has been shared.

3. Disposable income is gross recorded income for all household members less income tax paid. An alternative to disposable income for this type of analysis is equivalised household income. See the report for further discussion.

4. Cells containing “S” (in this and subsequent tables) were supressed due to StatsNZ’s confidentiality requirements. In most cases this means that fewer than 20 underlying datapoints contributed to the cell.

5. While long-term expenditure cannot exceed long-term income for a single household, this can be the case for a cross-section of households at a point in time. From a policy perspective, it would be useful to know the persistence of welfare dependence and low-income employment at a household level. That was beyond the scope of this project.

6. Cells between 95% and 105% are unshaded, as it could be misleading to draw strong conclusions about coverage ratios close to 100%.

7. Recent HES surveys asked two questions, each about different types of infrastructure services bills. Respondents can choose no difficulty, difficulty once, or difficulty two or more times. (‘Difficulty’ includes both shortage of money and late payment for any reason, including forgetting to pay on time.) In this analysis, a household has difficulty if they reported difficulty two or more times for either question, or they reported difficulty once for both questions.

8. See, for example, Anthony G Wilson, Ruth M Houghton & Rachel K Piper (1995). Budgeting Assistance and Low Income Families: Changes in Income and Expenditure. Social Policy Journal of New Zealand. Issue 05 (Dec 1995).