Using Household Economic Survey Data to Assess Food Expenditure Patterns and Trends in a Country with Notable Health Inequities


 Background Dietary factors are one of the most important risk factors for health loss worldwide, however there is limited information on dietary trends in Aotearoa New Zealand (NZ) and whether inequities in dietary patterns are changing.Methods We extracted data from the Household Economic Survey (HES), which was designed to provide information on impacts of policy-making in NZ. Three HES waves in 2006/07, 2009/10 and 2012/13 (n=9030) were used to examine the trends in household expenditure for key food groups by income-level and ethnicity, including food group expenditure (absolute amounts and as a proportion of total food expenditure, inflation adjusted).Results Overall, total household food expenditure per capita (CPI-adjusted) increased by 0.38% annually over this period (albeit not statistically significant). In absolute terms, low-income households spent around three quarters of what high-income households spent on food per capita. High-income households experienced a greater increase in expenditure on nuts and seeds (increasing the gap) and a greater reduction in expenditure on processed meat (reducing the gap because expenditure remained higher than in low-income households). There was increased expenditure over time on fruit and vegetables (NZ$273 per capita per year), nuts and seeds, and healthy foods (NZ$507) in Māori (Indigenous) households with little variations in non-Māori households. As a result, the relative health-related gap in food expenditure between Māori and non-Māori households declined over time. But there was little change in processed meat expenditure for Māori households and expenditure on less healthy foods also increased over time. Conclusions HES data were useful for understanding trends in food expenditure patterns, in the absence of nutrition survey data. Potentially positive expenditure trends for Māori were identified, which could indicate a shift towards healthier diets. However, food expenditure inequities in processed meat and less healthy foods by ethnicity and income continue to be substantial. Public policies that aim to support healthy diets for all could involve changes to the food environment and government support for culturally appropriate Māori-led interventions.


Abstract Background
Dietary factors are one of the most important risk factors for health loss worldwide, however there is limited information on dietary trends in Aotearoa New Zealand (NZ) and whether inequities in dietary patterns are changing.

Methods
We extracted data from the Household Economic Survey (HES), which was designed to provide information on impacts of policy-making in NZ. Three HES waves in 2006/07, 2009/10 and 2012/13 (n=9030) were used to examine the trends in household expenditure for key food groups by income-level and ethnicity, including food group expenditure (absolute amounts and as a proportion of total food expenditure, in ation adjusted).

Results
Overall, total household food expenditure per capita (CPI-adjusted) increased by 0.38% annually over this period (albeit not statistically signi cant). In absolute terms, low-income households spent around three quarters of what high-income households spent on food per capita. High-income households experienced a greater increase in expenditure on nuts and seeds (increasing the gap) and a greater reduction in expenditure on processed meat (reducing the gap because expenditure remained higher than in lowincome households).
There was increased expenditure over time on fruit and vegetables (NZ$273 per capita per year), nuts and seeds, and healthy foods (NZ$507) in Māori (Indigenous) households with little variations in non-Māori households. As a result, the relative health-related gap in food expenditure between Māori and non-Māori households declined over time. But there was little change in processed meat expenditure for Māori households and expenditure on less healthy foods also increased over time.

Introduction
Dietary risk factors are one of the most important risk factors for non-communicable diseases (NCDs) worldwide, accountable for 11 million premature deaths and the loss of 225 million disability-adjusted life years (DALYs) in 2017. (1) These risk factors contribute to cardiovascular diseases (CVD), diabetes, and cancer; which are among the top leading causes of deaths globally. (2) Household food expenditure surveys are increasingly used to monitor changes in dietary patterns internationally as individual nutrition survey data are often lacking and such nutrition surveys are expensive. (3)(4)(5)(6)(7) Employing food expenditure data, studies have suggested that urbanisation, industrialisation and globalisation among other factors have shifted dietary patterns towards more processed foods. (8)(9)(10)(11) Food-insecurity is also highly correlated with total household food expenditure, (12,13) and low-income households often lack access to nutritious foods. (14) In Aotearoa New Zealand (NZ), the NZ Burden of Disease Study shows that nutrition and obesity factors contribute to 18.6% of total health loss (in DALYs). In addition, diet and obesity related diseases are unequally distributed by ethnicity and deprivation, with Māori, Pasi ka and groups with low socioeconomic position at a higher risk of having obesity (15) and NCDs. (16) Much of this health loss and premature death could be prevented by improved diet and addressing the obesogenic environment that encourages unhealthy nutrition. Dietary patterns that are high in sodium, low in fruits and vegetables, low in nuts and seeds, high in processed meat and high in sugar-sweetened beverages are the major risk factors for NCDs including CVD, diabetes and cancer. (1,17) Despite the relative importance of dietary risk factors in generating health outcomes, the most recent Adult Nutrition Survey in NZ was over a decade ago (2008/09) and for children it was around two decades ago (2002). (18) For this reason, there are no recent representative data on trends in dietary patterns in NZ (but note that the NZ Health Survey included food frequency questions in 2019/20), nor changes in the distribution of diet by social factors.
Furthermore, different data sources have different strengths and weaknesses. Individual dietary intake includes many sources of food but often relies on the participants' memory. Trends in household food expenditure do not directly indicate individual diet, but are particularly relevant to concerns about food security and the cost of healthy food, particularly for low-income households (eg, the proportion of income spent on food). Expenditure data can also be used to examine changes in types of food bought over time.
The linkage of the NZ Household Economic Survey (HES) data (2006/07, 2009/10, 2012/13) (19) creates repeated cohorts of nationally representative data. The HES contains detailed information about household food expenditure, alcohol expenditure, tobacco expenditure, and other non-food household expenditure; and is implemented every three years. (20) HES can also contribute information on food consumption patterns. (4,21) We therefore aimed to explore the trends and social patterns in NZ household dietary expenditure using national representative linked HES data in NZ.

Results
There was a total of 9030 household survey participants evenly distributed across three HESs (2901 in 2006/07, 3126 in 2009/10, 3003 in 2012/13). Data on food expenditure was not available for 0.76% of households and they were excluded from the analysis; and 0.63% households had no income and so were unable to be included for the expenditure as a proportion of income outcome measure. Mean household income of the total sample was NZ$37,700.

Methods
We used data from three HES waves (2006/07, 2009/10, 2012/13) with a total of 9030 households. (19,22) These samples were randomly drawn from the total NZ resident population. The HES comprises questionnaires on the household, expenditure, income, and a two-week expenditure diary. Household expenditure includes: food (~500 food items), alcohol, tobacco, transportation, housing, health, education, recreation and culture, and other goods and services. There is a strong correlation between what individuals report about individual income in the HES and their actual income recorded in the Inland Revenue data. (22) Three speci c food groups were examined: fruit and vegetables (including frozen), nuts and seeds and processed meat, as per Global Burden of Disease (GBD) Study 2017 data, (1) Ni Mhurchu et al (2015) (23) which used HES data, and our previous work. (24) These food groups were selected because they are major risk factors for CVD (1) and can be categorised in the HES. An overall healthy food group was de ned using the nutrient-pro ling criteria as per Waterlander et al (25,26) and included all fresh fruit and vegetables (plus frozen fruit and vegetables), fresh seafood, nuts and seeds, whole grains, milk, legumes and bottled water. A less healthy group was de ned as all the remaining food and beverages, such as sugar-sweetened beverages, snack food such as potato chips, confectionary, and takeaway foods.
The main outcomes of interest for each food group were expenditure per person per year (2013 NZ$, annual values were provided by the data provider Stats NZ), expenditure as a proportion of total food expenditure, and expenditure relative to total income, all obtained from the HES data. We de ated income and expenditure to get comparable measurements across HES waves. We calculated mean expenditure and its standard error (SE) for total food and each food sub-group in each HES wave. Expenditure by average household income-level per person (high-and low-income were de ned as above and below the median for the HES survey respectively) and by household ethnicity (whether any household members self-identi ed as Māori [Indigenous population] or not) were also calculated. Expenditure trends, relative risks and signi cance levels were estimated using linear regressions, employing survey weights and adjusting for sampling structure. Independent variables for these linear regression models were survey year and either household income-level per person or household ethnicity.
Data were extracted using SQL version v.18.8, and were further processed and analysed in R using the 'survey' package, version R.3.6.0.
Māori households accounted for around 17% of the total sample (1,520 households). Māori households across all cohorts had signi cant differences to non-Māori households in income, age, household size and the number of households with children. Māori households had a lower mean income (NZ $30,400 per capita) compared to non-Māori (NZ$39,200 per capita). Māori households spent less on food per capita NZ$3,490 (11.5% of total income) compared to NZ$4,230 (10.8% of total income) for non-Māori.
Māori households tended to have more people, with a medium size of 3.09 compared to 2.43 for non-Māori and had a greater percentage of households with children; 48.8% compared to 34.5% non-Maori households with children (see Appendix A). For further characteristics of the survey samples see Table 1. Relatively, total food expenditure (out of all expenditure) appeared to increase by 1.13%/three years or 0.38% per year between 2006-2012. Expenditure on fruit and vegetables (-1.28% change), processed meat (-2.01%), and healthy food (-1.71%), appeared to decrease slightly over time. Expenditure on less healthy foods appeared to increase by 1.79%, and nuts and seeds by 5.85%. These estimated trends in expenditure had wide uncertainty (se >50% of the mean) and were not statistically signi cant.
Trends by income: Total annual food expenditure for low-income households appeared to increase by 3.11% to NZ$3,580 (se: 64.2) in 2012/13, whereas that gure for high-income households remained stable at NZ$4,840 (se: 97.7). However, there was a peak in low-income household spending in 2009/10, and at the same time a dip in high-income household spending, which are masked when considering linear trend results. Both types of households appeared to slightly reduce their expenditure on fruit and vegetables, and on healthy foods. Expenditure on nuts and seeds appeared to increase more in highincome (8.33%, se: 5.08) than low-income households (3.70%, se: 5.85). There was a reduction in highincome household expenditure on processed meat over the years by -4.20% (se: 2.37) not seen in lowincome households (0.00% change, se: 2.22).
Trends by ethnic group: Total food expenditure for Māori households increased by 7.82% to NZ$3,750 (se: 129.8) in 2012/13, but that for non-Māori households slightly reduced (albeit not statistically signi cant). Māori households increased their expenditure on fruit and vegetables (9.34%, se: 4.14), nuts and seeds (25.0%, se: 12.3), but there was little change in expenditure on processed meat (-1.05%, se: 3.72). Māori households appeared to increase spending on healthy foods (4.34%, se: 3.77%), whereas non-Māori households decreased their spending on healthy foods (-2.69%, se: 1.35%). Further details of these food expenditures are provided in Table 2. Note: a Numbers may not add up exactly as they were randomly rounded to meet con dentiality requirements. b Values in this column were derived using linear regressions with the survey year as the only independent variable, but no changes were statistically signi cant. All values in this Table were calculated or estimated using survey weights. Table 3 compares food expenditure by income-level and household ethnicity (relative risks) in each wave.
Low-income and Māori households spent less money on all food categories in all years compared to high-income and non-Māori households respectively, and almost all of these differences were statistically signi cant.
The gap in expenditure between low-and high-income households (where high-income households spend more) increased over time for nuts and seeds and decreased over time for processed meat. Changes were less clear for other food groups.
The gap in expenditure between Māori and non-Māori households appeared to decline over time for fruit and vegetables, nuts and seeds, and healthy food.  Proportions of household food group expenditure out of total food expenditure Table 4 presents proportion of speci c food group expenditure out of total household food expenditure by income-level and ethnicity for three HES waves. Overall, expenditure on fruit and vegetables accounted for 11% of the total food expenditure, less than 1% for nuts and seeds, 5% for processed meat, with 19% towards healthy foods and 81% for less healthy foods. There were small uctuations in the proportions of speci c food group expenditure out of total food expenditure, however, except for expenditure on nuts and seeds by Māori households (increased from 0.34% in 2006/07 to 0.49% in 2012/13), these changes were not statistically signi cant. Note: * , ** , *** Denote statistical signi cance at the 10%, 5% and 1% levels, respectively. Table 5 compares differences in the above expenditure proportions by income-level and ethnicity for each time-period. Low-income households spent greater proportions of their food budget on fruit and vegetables, processed meat and healthy foods than high-income households (peaking at 27% more than high-income households). Māori households spent greater proportions of the food budget on less healthy food especially in the rst time period (30% more on processed meat and around 5% more on less healthy food in 2006/07) compared to non-Māori households; and a lower proportion of the food budget on healthy food (68% fruit and vegetables, 49% nuts and seeds, and 80% for healthy foods in 2006/07 of the level in non-Māori households). These patterns largely persisted over the years, but with some nutritionfavourable trends for Māori households (eg, increased proportion on fruit and vegetables and decreased proportion on processed meat).

Discussion
This analysis identi ed some nutritionally-favourable expenditure trends for Māori. There was increased expenditure on fruit and vegetables, nuts and seeds, and on the healthy food category in Māori households, although expenditure also increased on less healthy foods as well. As a result, the relative gap in health-related food expenditure between Māori and non-Māori households declined over time.
However, a stark difference in expenditure remained, with around half to a quarter lower expenditure by Māori households in the healthy food groups. The trends in fruit and vegetables and nuts and seeds were similar for expenditure as a proportion of the food budget and as a proportion of income; although there was no signi cant change in these indicators for Māori expenditure on healthy foods. Some ndings were less favourable from a nutritional perspective. There was little change in processed meat expenditure for Māori households and it remained at a level just slightly less than non-Māori households.
Less healthy food expenditure as a proportion of total food expenditure changed little over time, but as a proportion of income it increased in Māori households more than non-Māori households (see Appendix B). Māori households differed signi cantly in income, household size and the percentage of households with children, so these factors may mediate the association between ethnicity and food expenditure.
Income inequities in food expenditure appeared to be relatively stable over time, however there were some potentially concerning trends from a nutrition perspective (unadjusted for ethnicity). High-income households experienced a greater increase in nuts and seeds expenditure (increasing the gap). Lowincome households had a greater increase in less healthy food expenditure, but it remained almost a third less than high-income households. Higher-income households also had a greater reduction in expenditure on processed meat (reducing the gap because expenditure remained higher than for low-income households).
Approximately 60% of the food available in NZ supermarkets have been classi ed as 'ultra-processed'. (27) In our study, we found roughly 81% of total food expenditure was spent on less healthy foods (  Table B1). NZ's grocery prices are also relatively high overall. In 2017 NZ ranked as the sixth highest priced grocery market in the OECD, (29) while its income per capita is below the OECD average. (30) New Zealanders also appear to spend a large proportion of their income on food and groceries. (29) The country's 'supermarket duopoly' may be a contributing factor and this currently being investigated by the relevant NZ Government agency: the Commerce Commission.

Strengths and limitations
This study was the rst (that we know of) to use HES data for informing food expenditure patterns in the NZ setting. Strengths of the HES data are that it is repeated every three years, is representative of the whole NZ population, and is a validated tool for informing economic policy. However, it should be noted that even though the Māori sample is at least large enough to perform the analysis, it does not provide equal explanatory power for Māori compared to non-Māori. These data included two-week diary for household expenditure so it reasonably covers the major food groups. However, there were some minor food category changes over the years in HES. There were also no data on non-commercial sources of food eg, home-grown, food from the wild (gathering kai moana [eg, shell sh], shing, and hunting), gifting of food (a feature of both Māori and Pasi ka cultures), and provision of free food via school breakfast and lunch programmes (for low-income schools). Nevertheless, these are not typically large sources of food for most people in NZ. Most importantly, there was no food purchase quantity data collected in this dataset and no accounting for food wastage (which can often be large in the fruit and vegetable category (31) ).

Research and policy implications
From a health and nutritional perspective there is a need to keep the HES and have a routine plan to analyse the food expenditure aspects. But it is also desirable to further contextualise the HES data with other routine data collection (eg, supermarket sales data -albeit somewhat expensive to purchase from commercial providers). Also having regular adult and child nutrition surveys would be even better.
Policy options that the NZ Government could consider to improve nutrition and reduce inequities in dietrelated diseases include: 1. (a) Introducing subsidies for healthy food (eg, provision of vouchers for purchasing discounted fruit and vegetables from farmers markets in low-income communities). (32); 2. (b) Making unhealthy food more expensive (eg, via food taxes (25) ). While such taxes could put more nancial burden on low-income households, this can be addressed by using tax revenue to subsidise healthy food and expanding food in school programmes. In addition, tax revenue could be used to subsidise farmers markets in more deprived areas.
3. (c) Increasing the regulation around the marketing of unhealthy food (eg, especially marketing children are exposed to).

(d)
Improving the nutrition-related labelling of foods and mandating warning labels on unhealthy foods (as used successfully in tobacco control, and being legalised in Chile (33) ).

(e)
Supporting culturally appropriate Māori led interventions to improve food security and healthy eating in Māori (34,35) including protecting wild-food resources from waterway and marine pollution.

Conclusions
In this study HES data were useful for understanding trends in food expenditure patterns, but limitations remain and further investment in nutrition survey data is recommended. There seems to be slow improvements in diet inequalities by ethnicity, and no evidence of any improvement by income, implying much more must be done to address nutrition to reduce the burden of NCDs and NCD-related inequities in this high-income country.