Stats NZ has a new website.

For new releases go to

As we transition to our new site, you'll still find some Stats NZ information here on this archive site.

  • Share this page to Facebook
  • Share this page to Twitter
  • Share this page to Google+
Technical notes


One useful way of comparing the incomes of different groups is to present the information in quintiles. Quintiles are created by ranking all incomes from lowest to highest and dividing the ranked incomes into five even groups. Each income is then assigned to a quintile and the distribution of the population subgroups across the quintiles can be compared. The lowest income quintile is usually referred to as quintile one, and the top quintile as quintile five.

Multivariate Analysis

The multivariate analysis technique used in this report was logistic regression analysis. This technique is often used to investigate the relationship between binary (yes/no) responses and a set of explanatory variables.

For the response variable Yes or No to the question about internet access availability in the dwelling, the stepwise method in SAS procedure PROC LOGISTIC introduces the explanatory variables in the order of their degree of influence on the response variable. It also produces Nagelkerke's adjusted general coefficient of determination. This gives an indication of the proportion of the variability of the responses explained by the explanatory variables similarly to that of classical regression.

The household characteristics included in the analysis were:  

  • total household income
  • highest qualification of the occupants
  • household composition by child dependency status
  • age of the youngest occupant
  • ethnicity of occupants
  • urban area classification
  • labour force status of the occupants
  • tenure of dwelling.

The set of dwellings for this analysis was pruned to exclude any with missing responses for total household income or highest qualification.

R2 – Nagelkerke's adjusted general coefficient of determination
Sequence of explanatory variables entered

as sole
explanatory variable

combined with preceding
explanatory variables
Total household income 19% 19%
Highest qualification of the occupants of the dwelling: none, school, vocational, degree 18% 25%
Dependent child household composition 11% 27%
Age group of youngest occupant (under 25, 26-59, 60 or over) 10% 28%
Ethnicity - presence of at least one occupant of Māori, Pacific, Asian or European ethnicity 3% 30%

The tenure, labour force status, and geographic variables all have statistically significant coefficients, but this is a feature of having a very large sample. Including these variables did not increase the adjusted coefficient of determination beyond 30 percent. Some interactions between the variables were tested, but were similarly ineffective.


Analysis of Internet access of census dwellings reveals several characteristics between whose sub-groupings there is a marked difference in the level of Internet access. Such characteristics include total household income, labour force status of the occupants, youngest age of the occupants, highest qualification of the occupants. Because labour force status is closely related to the total household income, a direct comparison of the Internet access of households by labour force status may be misleading. The difference may owe more to the fact that subgroups have different household income distributions than to the characteristic itself.

Standardisation is a technique commonly used by social scientists to remove the effect of age and sex from comparisons of subgroups of the population. The same method has been applied to compare Internet access once the effects of total household income have been removed from the subgroups of the population. The method of direct standardisation has been used, with the income distribution of the in-scope population of dwellings used as the base.

Because highest qualification of the household has also proved to be a useful explanatory variable for Internet access, for some variables highest qualification has been used as a base for alternative standardisation.

How the standardisation was applied

Direct standardisation involves looking at the cells of income groups within the characteristic group.

eg income group = $30,001 – $40,000
characteristic group = labour force status of part-time employment

The count of dwellings in the cell is adjusted to be at the same percentage within the characteristic group as within the whole population.

eg percentage of dwellings in income group
$30,001 – $40,000 from the population = 9%
count of dwelling with labour force status
of part-time employment = 106,631
adjusted count of dwellings in the cell = 9% of 106,631
= 9,597

The adjusted count of dwellings with Internet access in this cell is calculated by applying the rate of Internet access in that cell to the adjusted count

eg rate of access for the cell = 39%
adjusted count of dwellings with access = 39% of 9,597
= 3,754

The adjusted counts for the characteristic are summed across the income groups to obtain the adjusted count of dwellings with Internet access. From this the standardised proportion of dwellings with Internet access is obtained.

Formula to obtain the standardised count with an income group x characteristic group cell

Household Economic Survey

The New Zealand Household Economic Survey (HES) was conducted annually by Statistics New Zealand from 1 July 1973 until 1998 before moving to a three-year cycle. It switched to an April–March year from April 1975, but, for 2000/01, switched back to a July–June year. The HES provides a comprehensive range of statistics relating to income and expenditure.

Survey scope

The target population for the HES is New Zealand-resident, private households living in permanent private dwellings. This means that the population does not include overseas visitors who expect to be resident in New Zealand for less than 12 months; people living in institutions or in establishments such as boarding houses, hotels, motels, and hostels; homes for the elderly; hospitals; or psychiatric institutions. Inmates of penal institutions; members of the permanent armed forces; members of the non-New Zealand armed forces; and overseas diplomats are not included. However, children at boarding schools are accounted for by including expenditure on behalf of those children in the record-keeping of the parent or guardian.

For survey purposes, a household comprises a group of people who share a private dwelling and normally spend four or more nights a week in the household. They must share consumption of food, or contribute some portion of income towards the provision of essentials for living as a group.

Survey period

The survey was carried out over the period 1 July 2000 to 30 June 2001. People were asked about their spending in the 12 months preceding the interview.

Expenditure data was collected by the following methods:

  • 12-month recall (for single payments of $200 or more)
  • latest payment (for regular commitments, such as electricity, telephone, rates, rent, insurance and superannuation)
  • 14-day diary keeping.

Expenditure data collected by the diary covers a 12-month period (from 1 July 2000 for households interviewed in that month, to 30 June 2001 for those interviewed then). Expenditure data collected by recall in the expenditure questionnaire covers a two-year period (one year back from 1 July 2000 for households interviewed in that month, through to 30 June 2001 for households interviewed then). Reported expenditure has not been adjusted for the effects of that difference in coverage.

Similarly, for information on income, each household member aged 15 and over was asked about their income in the year prior to interview date. So income data covers a two-year period, depending on the month each household was interviewed.

Under-reporting of expenditure

For some items of expenditure, the total annual expenditure for all private households is less than that reported from other data sources. Some of the main reasons for this are:

  • Expenditure by residents of non-private households or by those ineligible for the survey (for example overseas visitors) is excluded from this survey.
  • Respondents to the survey forget or omit some types of purchases. This may include such items as cigarettes, alcoholic drinks, confectionery, newspapers and public transport fares.
  • Expenditure by children under 15 is not recorded in the survey.
  • There is a bias associated with non-response that affects some statistics.

No adjustments were made to the data to compensate for any under reporting. Items for which under reporting occurs in the HES are generally consistent with items that are under reported in similar overseas surveys.

Reliability of the survey estimates

The HES sample comprises 2,808 private households, sampled on a statistically representative basis from rural and urban areas throughout New Zealand. Information is obtained for each member of a sampled household that falls within the scope of the survey and meets survey coverage rules.

Two types of error are possible in estimates based on a sample survey – sampling error and non-sampling error.

Sampling error is a measure of the variability that occurs by chance because a sample rather than an entire population is surveyed. Relative sampling errors are calculated for average weekly expenditure and aggregate annual expenditure. For example, in 2000/01, the estimated average weekly household expenditure (excluding net capital outlay) was $758.30. This is subject to a percentage sampling error at the 95 percent confidence interval of plus or minus 3 percent. This means that there is a 95 percent likelihood that the true value lies between $735.60 and $781.00.

Non-sampling errors include errors arising from biases in the patterns of response and non-response, inaccuracies in reporting by respondents, and errors in the recording and coding of data. Statistics New Zealand endeavours to minimise the impact of these errors through the application of best-practice survey methods and the monitoring of known indicators (eg non-response). The overall response rate was 73 percent for the 2000/01 year.

New estimation methodology

A new estimation methodology, integrated weighting, was introduced for the 2001 HES. This is a recently developed method of adjusting the statistical output of a survey to match population person and household benchmarks. In particular, it takes account of undercoverage in the survey for specified population groups, such as young males and Māori.

Integrated weighting improves the robustness and accuracy of survey estimates. It also reduces the effect of bias in estimates resulting from undercoverage, as well as reducing the level of sampling error for benchmark variables. It permits the calculation of one weight for each household that can be used for both individual and household estimates.

  • Share this page to Facebook
  • Share this page to Twitter
  • Share this page to Google+
  • Share this page to Facebook
  • Share this page to Twitter
  • Share this page to Google+