Stats NZ has a new website.

For new releases go to

www.stats.govt.nz

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+
Post-enumeration Survey: 2013
Embargoed until 10:45am  –  27 May 2014
Commentary

National net census undercount of 2.4 percent

Overall, coverage in the 2013 Census was high, reflecting the cooperation and support of the New Zealand public, and the high-quality work of field staff. The 2013 Post-enumeration Survey (PES) showed that about 97.6 percent of New Zealand residents in the country on census night were counted in the 2013 Census. This means there was a net undercount of 2.4 percent, or 103,800 people. The net undercount figure results from an estimated gross undercount of 135,500 people (3.1 percent), offset by 31,700 people (0.7 percent) being counted more than once (overcount). See table 1 for detailed results. 

Given the small size of the sample, the net undercount figures provided by the 2013 PES are subject to sizeable error margins, called the sampling error. Therefore, the undercount estimates need to be interpreted with caution. The overall sampling error was 0.5 percent. This means we are 95 percent confident that the undercount lies between 1.9 and 2.9 percent, not allowing for non-sampling error.

International comparison

Under-enumeration is a phenomenon shared by many census-taking nations. New Zealand's net undercount rate is comparable with the most recent rates available for other countries. For example, the net undercount for the 2011 Canadian census was 2.3 percent, and for the 2011 Australian census it was 1.7 percent. The methodology used in each country differs in many important ways, so treat international comparisons with caution. 

Sex and age: Undercount higher for males and young adults

There were variations in net undercount by sex and age. The 2013 Census missed more males than females, as did the three previous censuses. With a net undercount of 2.6 percent for males and 2.1 percent for females, the difference of 0.5 percentage points between the sexes was smaller than those in 1996 (0.6 percentage points) and 2001 (0.7 percentage points), but larger than in 2006 (0.3 percentage points).

Young adults (aged 15–29 years) had the highest net undercount, of 4.8 percent. Children (aged 0–14 years) had an undercount rate of 3.0 percent, well below that for young adults. New Zealanders aged 30 years and over were well enumerated, with net undercount rate of 2.0 percent for those aged 30–44 years, 1.3 percent for those aged 45–64 years, and just 0.5 percent for those 65 years of age and older. The relatively high level of undercount for young adults is shared with other comparable countries.

Graph, Net undercount by age, 2013 Census.

Ethnic groups

Undercount continues to vary between ethnic groupings. Care is required when comparing the ethnic undercount rates with other undercount rates in this information release and with previous PESs, because of methodological differences. The rates given in this section, however, are comparable with each other. This enables us to understand the relative undercount for each ethnic grouping.

The net undercounts by ethnicity were:

  • Māori – 6.1 percent
  • Pacific – 4.8 percent
  • Asian – 3.0 percent
  • European – 1.9 percent.

The major change between 2006 and 2013 was among the Asian population, with the undercount halved. This may be related to the drop in the number of international students over the period. International students are frequently missed by censuses or miscoded as overseas residents. Also, being predominantly in the young adult age groups, students have a higher likelihood of not being counted in the census. Sampling errors were high for people of Asian ethnicity in 2006, so treat the 2006 result for this group with caution.

Graph, Net undercount by ethnicity, 2013 Census.

Geographic area: Undercount highest in northern North Island

The level of undercount was highest in the northern half of the North Island, at 3.0 percent. The northern North Island is made up of the highly diverse regions of Northland, Auckland, Waikato, and Bay of Plenty. This compares with the undercounts for the rest of the North Island at just 1.3 percent, and the South Island at 2.2 percent.

Differences in undercount by area can be partly attributed to the ethnic and age composition of the population, notably in the case of the northern North Island. The northern North Island not only contains over half of the total New Zealand population, it also has a higher proportion of its population under 30 years of age, at 42 percent, compared with the rest of the North Island (40 percent) and the South Island (38 percent).

Among the major population centres within these broad regions:

  • Auckland region had a net undercount of 2.1 percent
  • Wellington region had a very small net undercount, at 0.7 percent
  • Canterbury region had a net undercount of 2.9 percent.

The undercount for the rest of the North Island, excluding Auckland and Wellington regions, was 3.4 percent, compared with 1.3 percent for the rest of the South Island excluding Canterbury.

Graph, Net undercount by area, 2013 Census.

Response rate

The response rate is the proportion of eligible households who responded to the PES interviewers. A responding household was one where there was at least one fully responding eligible person. The 2013 PES had a response rate of 87 percent, which was below the target rate of 90 percent, but is comparable with other household surveys conducted by Statistics NZ.

Census targets

The 2013 Census had established a set of indicators to measure the success of census enumeration and processing. Census did not meet either the coverage or the response rate targets. PES estimated that the national coverage rate of 97.6 percent was just below the target for coverage of 98.0 percent. The PES also found that the achieved census national response rate of 92.9 percent was below the target of 95.0 percent. 

Reasons for census undercount and overcount

Reasons for undercount (people and dwellings being missed) in the census include newborn babies being overlooked, people shifting from one house to another around census time or deliberately avoiding the collector, occupied dwellings misclassified as unoccupied, and dwellings missed by collectors.

Situations where overcount can occur include children living in joint custody with both parents completing forms for the same child, or children attending boarding school, people with dual residences, unoccupied dwellings misclassified as occupied, and residents who are temporarily overseas on census night.

Census substitute records and PES

The methodology used for the 2013 PES enabled a partial investigation of the substitute records raised by the 2013 Census. Previous PESs assumed that the number of substitutes created was correct. This was recognised as a missed opportunity with implications for the interpretation of census coverage. So long as the number of substitute records remained relatively small in census, this had been a less significant problem. However, a total of over 203,000 substitute records were raised in the 2013 Census.

This was more than in previous years. Of these, 135,000 were in private households where all members and the dwelling form were substitutes (referred to as 'an all-substitute dwelling'). For the first time, the 2013 PES measured the validity of this process in cases where the PES recorded a match to both a substitute record in an all-substitute dwelling and a record found elsewhere. This allowed us to identify and adjust for an incorrectly raised substitute form. The 2013 PES found that the 2013 Census raised approximately 15,000 too many substitute records of this type.

Undercount estimates for ethnic groups were not adjusted in this way because we do not impute ethnicity in census.

Understanding substitution and imputation in the 2013 Census has more information about substitutes and imputation of key variables.

Non-response to 2013 Census was 7.1 percent

The combination of net undercount, estimated by PES, and substitute records, counted in census, give a more complete measure of non-response to census. In 2013, the non-response to census was 7.1 percent of the PES estimated resident population on census night. In 2006, the non-response to census was 5.2 percent. In 2001, it was 5.0 percent. While the undercount has remained between 2.0 percent and 2.5 percent over this time, the number of substitute records has increased, resulting in higher non-response.

PES helps form the base for post-censal population estimates and projections

The 2013 Census counts will not be revised in light of the 2013 PES results. However, the 2013 PES results are a key part of the 2013-base estimated resident population (ERP). The ERP 2013-base uses the 2013 Census and PES results, together with adjustments for residents temporarily overseas on census night and components of change (births, deaths, and migration) between census date and 30 June 2013. The ERP 2013-base is the starting point for deriving population estimates and projections, until a new base is derived after the next census.

The 2013-base estimated resident population will be released on 14 August 2014. At this time the population estimates for June 2013 to March 2014 quarters, currently based on the 2006 Census, will be superseded by 2013-base estimates. Population projections using the new 2013 base will follow from November 2014.

For more detailed data from the Post-enumeration Survey see the Excel tables in the 'Downloads' box.

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