Maintaining knowledge of the New Zealand Census *

1 of 8 21/08/2007 2:21 PM Symposium 2001/25 20 July 2001 Symposium on Global Review of 2000 Round of Population and Housing Censuses: Mid-Decade Assessment and Future Prospects Statistics Division Department of Economic and Social Affairs United Nations Secretariat New York, 7-10 August 2001 English only Maintaining knowledge of the New Zealand Census * Frank Nolan **

2 of 8 21/08/2007 2:21 PM CONTENTS A. Introduction B. Maintaining census capability C. Structure D. Staff retention E. Documentation F. Outputs and data G. Conclusion

3 of 8 21/08/2007 2:21 PM A. Introduction 1. The population census provides a cornerstone for the population and social statistics system of a country. It is often the means for allocation of considerable public funding, especially in areas such as health, education and social policy. The census can be used for electoral redistribution. 2. Maintaining census capability during intercensal years is important as a risk-management strategy. It can help to ensure that census information continues to be available to the required quality standard. The capability is needed both with staff and with documentation. This paper looks at the particular circumstances of the New Zealand census, and the operation of the census cycle. 3. New Zealand's 31st census of population and dwellings was taken on Tuesday, 6 March 2001. The census is a statutory requirement under the Statistics Act of 1975. The census is taken every five years. The 2001 census was conducted using a self-completed questionnaire that was delivered and collected from every household in the country. It is completed as a de facto count, recording where people are on census night. 4. For the 2001 census, two types of questionnaires were used: an individual form for every person in the country on census night, and a separate dwelling form for every dwelling occupied on census night. The individual questionnaire contained 43 questions covering topics on demographic and social structures, education, ethnicity and culture, geography, disability, income and work. The dwelling questionnaire contained 22 questions covering topics on housing, transport and household communications. The topics were similar to those used in the 1996 census. In addition, both questionnaires were available in the English language and as a bilingual form in Maori and English. 5. The enumeration was carried out by 6,500 field staff, most of whom were employed for a five-week period around census day to deliver and collect census forms. The field team was supported by a full public relations programme. This included advertising through all media channels (with special emphasis given to television), involvement with a range of community groups, a programme for schools and employing local community members as staff. 6. Processing of census questionnaires started on 27 March. The processing system built on the 1996 developments. The 1996 system consisted of imaging the forms, followed by recognition of mark boxes and numbers. This was followed with key-entry operators completing work where the recognition software was unable to code a response. There was extensive micro-editing, followed by output editing. 7. For 2001 there have been technological advances in imaging. The result is that only answer boxes are retained rather than the whole questionnaire, which reduces the volume of imaged data by 90 per cent and allows all images to be held on one server rather than being cut to CD-ROMs. Higher recognition rates have been achieved and some alphabetic character fields recognized. Micro-editing is limited, with the main check on the data being through output editing. 8. The first provisional count figures, released on 28 May 2001, gave New Zealand a census night count of 3,792,654 persons and 1,357,983 occupied dwellings. Final count data (on a usual-residence basis) is planned to be available before the end of December 2001, and full data are expected to be available before the end of February 2002. Data for electoral redistribution is required by 17 September 2001. 9. Considerable information is publicly available from the Statistics New Zealand web site. This includes national and regional counts published through media releases. A series of topic reports (covering such areas as education, ethnicity, housing and employment) is produced, consisting of detailed tables. These are both on the web and printed. Profiles for local authority areas will be produced, along with a set of finer community profiles. This provides information at a very fine level (slightly larger than enumerator district) as a return to communities for their participation in the census. 10. Three priced products are available. Statistics New Zealand staff can produce customized information for clients (generally tables at fine levels or using non-standard geographies). Researchers may access the anonymized census records in a data laboratory. They are available from Statistics New Zealand offices but are under strict controls both on access and on release of information.

11. A further fine-level tabular product has been provided in the last two censuses using Space Time Research's Supermap. For 2001, this will be replaced by a web tabulator using the Beyond 20/20 software. 12. The 1996 census of population had a net coverage of 98.8 per cent, as measured by a post-enumeration survey. B. Maintaining census capability 13. The census can be viewed as an ongoing continuous survey, albeit with a longer periodicity than most ongoing surveys. The five-yearly cycle in New Zealand means that there are advantages for maintaining capability compared with countries with a ten-yearly cycle. However, similar principles should still apply. 14. The first requirement is to recognize that, as with any ongoing survey, a key element in all considerations is consistency. This will be critical in measuring change between survey periods. So decisions to change census systems, often because of relevance or technological advance, need to be carefully balanced and managed with regard to the possible impacts on the results. Change management becomes very important: What will be the impact of changes on the data? The census does not offer the possibility of a dual run to measure change due to different design. 15. The census is often the largest survey undertaken by a national statistical office. As a consequence, it has a large budget. This generally means that there is an opportunity for funding many developments that may not have been otherwise undertaken by the office. There will also be large budgets for fieldwork, processing and output. Using this funding across multiple years and following the survey cycle offers the possibility of rotating staff and so maintaining knowledge of census processes. 16. Institutional memory can also be enhanced by having an organizational structure that contains a census division. This provides a home for maintaining census documentation, administrative processes (especially planning and finance), staff and unit-record databases across many census cycles. C. Structure 17. A census division is important for the retention of institutional knowledge. In 1996 Statistics New Zealand attempted to operate the census across a functional structure using matrix management [1]. This proved problematic. In development, the major shortfall was having no one manager accountable for the outcomes. The result was decisions by consensus, with everyone getting what they wanted. And it was hard for the different functions to see or focus on the end product. 18. For the 2001 census Statistics New Zealand reverted to the subject-matter structure for the census (for the most part). So the functions were generally under the one division, and there was accountability in one place for the outcome of the work. 19. Accountability is important, as the census can be the largest public undertaking of the statistical office. All parts need to be carefully managed to ensure that they make a complete whole. This is especially the case in managing the risk (which is significant) and the quality. 20. A key element in structure is planning. Very early in the census cycle the strategic approach needs to be agreed upon. This is important, as it sets the direction for the work. This can be followed by detailed plans setting the milestones for each project, along with the quality criteria and risk-management strategies. 21. A key section in this work is the coordination section. This group provides an overview for planning, finance and some administration matters. It provides a base for continuity between censuses, with the responsibility for ensuring that key documentation is archived and knowledge of the programme plans and budget information retained. 22. A vital part of the work of the coordination section is drawing together costing to obtain funding for the next census. Most of this work is in the intercensal period. There must also be a detailed understanding of how different activities will use the budget. There needs to be an ongoing process of managing funding (from government) and reporting on expenditure against budget. In New Zealand the budget is close to 15 per cent of the statistical office s budget across five years and about 30 per cent in census year. 4 of 8 21/08/2007 2:21 PM

5 of 8 21/08/2007 2:21 PM 23. A challenge to the capability is having some census functions contracted outside the statistical office. In New Zealand there has been an increased use of external contractors for the 2001 census to deliver such non-core elements as maps, call centre and imaging. 24. Increased use of external contractors does have the risk of loss of knowledge of the activity. However, good contract documentation does provide information for subsequent censuses. Critical to this is high quality and ongoing evaluation of the contracts. The 2001 census structure is shown below. CENSUS MANAGER Content and Output (including questionnaire design) Census Processes (development of enumeration, processing and output systems) Enumeration Communication Output Post-Enumeration Survey Coordination D. Staff retention 25. The census is a large, high-profile survey. It often attracts staff who are retained through several census cycles. This is partly because of the range of work that can be provided, as census staff are able to work through different census phases. 26. Following the 1996 census, permanent staff were placed in one of three sections: content and output, development and administration. Three temporary sections were established for enumeration, communication and processing. 27. While some staff have remained in one section, others have moved from one work area to another, depending on the cycle. Staff from content and output have a key part in first understanding the requirements of data users, and then delivering those outputs to users. In a five-year cycle this is an ongoing process. As soon as content development is completed, there is work on questionnaire design. Then it is time to start developing outputs and delivering outputs, followed by content development again. So a core of staff working in these processes retains information critical to the success of the census. 28. Staff of the development section have been used in the temporary operational sections. Staff can move from developing the enumeration systems in the operation of enumeration (and likewise in processing and outputs). This can provide good balance, as different phases require different staff levels on each project. 29. Following the operational phase of the census, those staff who have been working in operational areas can return to the development section to work on improving systems for the next census. 30. The activity peaks in the census often mean that there are opportunities for many staff of the statistics office to become involved in the work of the census. This can spread knowledge within the organization. The census also provides an opportunity for staff in the office to increase their skills with temporary work on census projects. There is also an opportunity to evaluate the skills of temporary staff, who may subsequently be further employed in the office.

6 of 8 21/08/2007 2:21 PM 31. The high profile of the census also means that most senior staff of the national office are at least involved in the strategic plans for census. Certainly the members of the executive staff are often involved in the communication elements during operations around census day. 32. The most difficult area in which to retain knowledge is in the field team. With such a large team it is impossible to retain them between censuses. However, a number do return from one census to the next. A number of the field staff do enjoy the short duration and public-contact work and also participate in other similar projects, like being election officials. 33. As a means of assisting in the development of enumeration processes, a plan has been developed for a small number of the 2001 enumeration managers to be retained part time to assist in reviewing the development of the 2006 enumeration process. Further staff will be established in local communities to assist in local capability-building as part of a wider government initiative. Both of these moves will assist in retaining knowledge from one census to another, especially at the local level. 34. In operational areas census knowledge can be maintained by making good use of succession planning. There are real opportunities for staff from one census to move up to a management level in the next census. This has been important in maintaining quality in both the enumeration and processing operations. E. Documentation 35. In New Zealand census documentation has traditionally been stored in a series of paper-based files. These have broadly covered administrative matters, enumeration activity, communications or public relations, questionnaire design, processing systems and outputs. 36. Since 1996 the census documentation has for the most part been written electronically and stored in a computer. The work for the 2001 census used a Lotus Notes database for all planning and discussion. 37. Three types of documentation are useful: plans of what was to be done; commentary on what actually was done; and an evaluation, especially looking at what should be done the next time. The best documentation is generally found in the first and last types of documentation. Often the activities of the day crowd out documenting what is taking place. 38. A particularly important element of documentation is the specification for tendered work and that for computer programs used, especially with survey processing. The tender documents need to be retained during the life of the tender to ensure that the specifications are being met. But at the completion of the contract, an evaluation needs to look at improvements that could be made to the specifications to improve the outcomes. This documentation can then be used for the next census. 39. For a census with a long periodicity, the specification for computer programs (mainly for data processing) may be more useful than the programs themselves, given the speed of technology change. 40. Planning information is vital for managing a project of the size of a census. Building on documentation from prior censuses assists in improving the quality of the work and reducing the risk to the project. Reviewing of documentation of the previous census project plans is important for understanding critical timing issues. It also assists in understanding allocation of resources to achieve the required project outputs. 41. Bringing together the start and end of the census process is also critical in maintaining good documentation. Having started by assessing the requirements of those who want to use information from the census, it is then important to deliver those data back to those groups. With the time frame for a census, this may mean user requirements are determined two to three years prior to census day, but output is delivered a year after. So there is a gap of three to four years. F. Outputs and data 42. A key outcome of any survey is the statistical information that is produced. Part of the process of maintaining capability is continuing to have access to these data.

7 of 8 21/08/2007 2:21 PM 43. Traditionally the data have appeared in printed publications produced mainly as tables. These publications have appeared from the time of the first census (1851 in New Zealand). Generally early publications have considerable geographic detail but are limited in the level of cross-tabulations. And there is no ability to retabulate the data. 44. More recent data (since 1976 in New Zealand) are securely stored as a unit-record database. This now offers the option of ongoing production to satisfy new needs with old data, often to measure change over time. 45. However, there is still an ongoing requirement to maintain these archival data in a medium which can continue to be used. This may entail conversion into a new software product (in New Zealand all data were moved from mainframe machines to a client server base in 1998-99). 46. Equally important is maintaining knowledge of the metadata. This is true not only for computerized data, but also for those tables that are available only in hard copy. At the most basic level, this means perhaps the questionnaire and some information on the methods of enumeration and processing. This is generally the limit for very early censuses. 47. For more recent unit-record databases, documentation includes a comprehensive description of the variables being stored and the classifications being used at the time (there may be a later requirement to recode the data to a newer version of the classification). It should also include information on the survey design and outliers in the data. This may appear in a Sources and Methods document. Not only is this worthwhile for users of census information, but it is also a good summary for future developers. 48. Having the data and the documentation in databases and software that are common in the statistical office also assists in ensuring capability. So staff who have knowledge of the database or software can work across a range of information. And when these tools are updated, the census data can also be updated. 49. Delivery of the outputs continues well after the census data have been produced. Again, there is a need for staff with knowledge of census data (over several censuses) to be available to assist in delivering information to clients. There is a need to maintain this knowledge in the statistical office over the long term. G. Conclusion 50. A census of population and housing is normally the biggest survey undertaken by a national statistical office. All operations generally exceed the usual survey cycle process by several orders of magnitude. Thus the census cycle is generally at least a five-year period. 51. Maintaining capability over the cycle is a difficult yet necessary requirement for the statistical office. It is difficult, since most census staff will not be able to be employed continuously over the census cycle. However, it is vital in that a key measure of the census will be change over time, and there needs to be careful management of changes to the survey process so as to understand which changes may be due to different processes and which are real-world changes in society. 52. Institutional memory in an organization is often dependent on ability to retain staff. This may be strong in a census team, given the high profile of the work. Complementing this will be thorough documentation of processes to ensure that change is carefully managed. In addition, a key capability to maintain is understanding and access to past census data. This is critical to understanding change over time. Unit-record data access is important to evaluate new trends from future data. 53. Census capability involves risk management to ensure that future censuses can be conducted to an increasing level of quality. * ** This document was reproduced without formal editing Statistics New Zealand, New Zealand. The views expressed in the paper are those of the authors and do not imply the expression of any opinion on the part of the United Nations Secretariat

8 of 8 21/08/2007 2:21 PM [1] In matrix management an individual has two lines of responsibility: the subject matter and the function. So the responsibility for the final product is shared among a range of managers who are responsible for different components of the product.