This page is not a current Christchurch City Council document. Please read our disclaimer.
| Year |
|||||||
| 1997 |
1996 |
1995 |
1994 |
1993 |
1992 |
1991 |
|
| Total addresses drawn for sample |
1,168 |
1,167 |
1,167 |
1,165 |
1,105 |
1,097 |
1,146 |
| Response rate1 |
78 % |
75 % |
72 % |
79 % |
82 % |
80 % |
73 % |
| Ineligible addresses |
132 |
116 |
119 |
110 |
79 |
102 |
98 |
| Total participating in survey |
803 |
789 |
752 |
826 |
833 |
792 |
754 |
| Non-contacts |
53 |
117 |
84 |
74 |
52 |
49 |
156 |
| Refusals |
180 |
145 |
212 |
155 |
141 |
154 |
138 |
1 Response rate = eligible responding
eligible responding + refusals + proportion of non-contact households who are eligible
Notes:
(i) If a household or household members did not meet the scope and coverage rules they were classified as ineligible.
(ii) If a household could not be contacted (as a rule three attempts were made) and there was no evidence that they were absent for the whole survey period, the address was classified as a non-contact.
Two types of error are possible in an estimate based on a sample survey: non-sampling error and sampling error. For any survey, possible sources of non-sampling error include inaccuracies in recall or reporting by respondents, error in the recording of data, and errors made in the processing of data.
Steps taken to minimise and control non-sampling error included:
Due to changes in wording of core questions between 1994, 1995 and 1996, and changes to the order of some questions over the same period, the user should be wary when making comparisons between these years.
Sampling error is a measure of the variability that occurs by chance because a sample, rather than a whole population, is surveyed. The size of the sampling error is controlled by the size of the sample and depends on the variability of the target population. To ensure the method is consistent with the theory used to calculate the sampling error, random selection and sample allocation techniques must be employed.
As a multi-stage cluster design was used, the sample errors may not be the same even for questions with the same number of respondents. This is due to the correlation of responses within clusters (intracluster correlation).
The sample error for the surveys remains in the order of 3% at the 90% confidence interval for those questions applicable to all respondents.
It should be noted that the 1991 - 1993 surveys were conducted during February, whereas the 1994 - 1997 surveys were conducted during May. This may have a significant effect on ability to compare data between 1991 - 1993 surveys and 1994 - 1997 surveys.
Changes in wording of questions as well as changes in the order of the questions also may have a significant effect on data comparability between survey years. In the 1991 - 1993 surveys the first and main part of the questionnaire stayed unchanged to allow comparability between years. The user should be wary when making comparisons when these kind of changes have been made.
The estimated sample errors for sub-populations in the survey are given in the table on page 12. While given nominally at a 90% confidence level, care must be taken for small sub-populations due to possible biases with the commensurately small samples achieved in the survey:
| Size of Sub-Sample |
Estimated Sample Error |
| 441-660 |
5% |
| 316-440 |
6% |
| 236-315 |
7% |
| 186-235 |
8% |
| 151-185 |
9% |
| 121-150 |
10% |
| 101-120 |
11% |
There was no imputation of individual responses, thus where a response was omitted, it was assigned to the "don't know" category. The standard assumption that those who refused are similar to those who participated in the survey was made.
Since the survey is a multi-stage cluster design, the details of the estimation procedure and sampling errors are not given here. However, Statistics New Zealand has used a ratio estimator calculation. Details can be found in W.G. Cochran, Sampling Techniques, 3rd edition, 1977.