Volume 1

Contents

1. Multiply Imputed Synthetic Data Files
2. Uniques and Disclosure Control Design
3. Methods for Constructing Spatial Cost of Living Indexes
4. Investigation of New Zealand's Inward foreign affiliates trade statistics using external data sources
5. Ethnic intermarriage in New Zealand
6. Ethnic intermarriage and ethnic transference amongst Māori
7. Feasibility Study: A Review of Selective Editing
8. Feasibility of using samples of telephone numbers for Tier 1 Official Statistical Household Surveys
9. Generating synthetic unit-record data from published marginal tables

The contents of these files are in Adobe Acrobat Reader format. If you do not have the Adobe Acrobat Reader you may download the reader to view or print them.

1. Multiply Imputed Synthetic Data Files (PDF, 367KB)
Patrick Graham, Richard Penny

Abstract: In this paper we review the Bayesian logic which underpins a proposal of Rubin to apply multiple imputation to the problem of constructing synthetic versions of survey datasets which official statistics agencies can release to researchers without compromising the confidentiality of the survey responses.

We investigate the application of the multiple imputation paradigm to constructing synthetic versions of a small dimensional but real dataset. We emphasise the use of hierarchical Bayesian imputation models in order to reduce the dependence of synthetic datasets on special structural model forms.

In our example, analyses of synthetic data generated under hierarchical Bayesian models exhibit more robustness to the imputation model than do analyses of synthetic data based on non-hierarchical versions of the same imputation models. However, so long as the analytical model is simpler than the imputation model, the impact of the imputation model on inference appears slight.

The confidentiality aspects of synthetic data are considered in the context of our example and we conclude that disclosure risks are no greater, and probably less, than those associated with the established confidentialising technique of random rounding.

Keywords: Bayesian modelling; Synthetic data; Multiple imputation; Statistical disclosure limitation.

2. Uniques and Disclosure Control Design (PDF, 384KB)
L. Fraser Jackson

Abstract: The statistical disclosure control (SDC) literature has traditionally focused on tools to prevent disclosure once a potential disclosure problem has occurred with little attention to describing the circumstances which generate sensitive cells and ways of avoiding them.

The objective of this paper is to study the factors which lead to numbers of sensitive cells and provide some guidance on rules which help reduce those numbers and the reliance which must be placed on SDC to preserve the anonymity of data providers.

Keywords: statistical disclosure control; confidentiality; sensitive cells.

Back to top

3. Methods for Constructing Spatial Cost of Living Indexes (PDF, 645KB)
Daniel Melser, Robert Hill

Abstract: This report, commissioned by Statistics NZ, investigates the construction of spatial cost of living indexes for New Zealand. The report address the conceptual and methodological, as well as practical, issues concerning the construction of spatial price indexes.

The report begins by outlining its motivation and goals and then sets out the various conceptual approaches to price index construction; the cost-of-living and cost-of-goods perspectives as well as the acquisitions, uses and payments approaches to index construction. These are discussed in the somewhat novel context of spatial as opposed to temporal price comparisons. Some particularly problematic headings both conceptually and practically such as housing are discussed at length.

The report outlines international practice in the construction of spatial price indexes which may be of some use to Statistics NZ. The computation of spatial price requires the choice of an index number formula, a multilateral method, an approach to aggregation, and the choice of weights amongst other decisions. These choices are discussed in detail and some guidance is provided on what the most appropriate method may be for Statistics NZ.

The authors take a look at the data currently available to Statistics NZ in the various product groups and discuss how this might be used, and what additional data may be required, in the construction of spatial price indexes. The final section contains some conclusions and recommendations.

Keywords: Spatial cost-of-living, price index, quality change, multilateral methods, EKS.

4. Investigation of New Zealand's Inward foreign affiliates trade statistics using external data sources (PDF, 193KB)
Jason Attewell, Wido van Lijf

Abstract: Foreign Affiliate Trade Statistics (FATS) can provide useful measures of the impact of foreign direct investment on an economy. Foreign affiliates are entities with overseas ownership greater than 50 percent, and are often established to deliver goods and services that require close contact between the supplier and the consumer. However, because foreign affiliates are considered resident entities in their host country, their transactions in these countries are not recorded in the Balance of Payments and International Investment Position statistics, but they do form part of the host economy's Gross Domestic Product (GDP).

FATS statistics attempt to measure the impact of foreign investment on the host economy and supplement existing statistical areas such as foreign direct investment and international trade in goods and services.

This paper looks at New Zealand's inward FATS, which are statistics relating to the activities of foreign-owned enterprises in New Zealand. The paper examines how the contribution of these firms to the New Zealand economy can be determined using existing data sources. It also touches on how and where New Zealand's inward FATS might be used to support policy research and other areas of enquiry.

Keywords: Foreign Affiliate Trade Statistics (FATS); foreign-owned enterprises in New Zealand; foreign investment; New Zealand economy.

Back to top

5. Ethnic intermarriage in New Zealand (PDF, 616KB)
Paul Callister, Robert Didham, Deborah Potter

Abstract: One of the recommendations of Statistics New Zealand’s 2004 Review of the Measurement of Ethnicity was that a programme of research on ethnicity be commenced. One of the central issues considered in the review was how best to collect, code and report dual and multiple ethnic responses. While the review recommended more use be made of single and combination ethnic groups, as well as total ethnic counts, there was still some uncertainty over the best way to report and use dual and ethnic responses.

This paper provides some practical samples of the presentation of multiple response data. It is hoped these examples will stimulate further analysis which acknowledges the complexity of ethnicity. Both birth registration data and census data indicate that a small, but growing, proportion of the population report more than one ethnic group. The reportage of more than one ethnic group is particularly prevalent amongst the wider Māori and Pacific Peoples ethnic groups, but is also important amongst younger European and Asian peoples.

Keywords: Ethnicity; Intermarriage; multiple response data

6. Ethnic intermarriage and ethnic transference amongst Māori (PDF, 149KB)
Sarah Howard, Robert Didham

Abstract: The people of New Zealand have a long tradition of ethnic intermarriage and as participants in this trend Māori have a unique population that has undergone enormous demographic and social change.

Māori have and continue to experience high rates of intermarriage, this along with adaptation to other social phenomenon has caused the measurement of ethnic identity to grow increasingly complex. The Māori ethnic group, rather than being homogenous, consists of many individuals from varying backgrounds who have varying cultural values, norms and identities but nevertheless at some degree choose to identify as being Māori. There has been widespread public interest and growing political debate surrounding the questions "Who is Māori?" and "How do we accurately define Māori?" (Kukutai 2001, Pool 2001, Pool 1963)

Keywords: Ethnicity; Intermarriage; multiple response data; Māori.

Back to top

7. Feasibility Study: A Review of Selective Editing (PDF, 364KB)
Carl Scarrott

Abstract: This report forms part of a feasibility study towards the implementation of selective editing across Statistics NZ and the Official Statistics System, under an OSRDAC research programme. It provides a review of the published literature on selective editing methodologies and key aspects of current international best practice.

The key features and cost benefits of the approach are detailed along with discussion of some implementation issues. As a preparatory stage for the development of an implementation strategy, this review goes on to outline a broad selective editing framework which is intended to be general enough to apply across Statistics NZ surveys. A number of recommendations have been made based on the evidence provided in the literature review, along with some suggested foundational and research questions for future consideration.

Keywords: selective editing, quality improvement, performance monitoring

8. Feasibility of using samples of telephone numbers for Tier 1 Official Statistical Household Surveys (PDF, 274KB)
Alistair Gray, Ramya Suntheralingam

Abstract: This paper updates Statistics New Zealand and the New Zealand public sector of how changes in telecommunications technology or the telecommunications environment impact on sampling individuals or households using telephone numbers as the sampling frame.

A major objective is to provide sufficient information to decide for Tier 1 Official Statistics whether using telephone numbers (fixed or mobile) as a sampling frame for individuals or households is feasible, either as stand alone frame or as part of a multiple frame approach.

It is not the intention of this paper to provide a manual outlining how to construct a telephone sampling frame or get access to it or how to sample from it. However, some information relevant to these issues will of necessity be provided in this paper.

Keywords: Telephone surveys; Sample design, Non-response, Coverage bias.

Back to top

9. Generating synthetic unit-record data from published marginal tables (PDF, 515KB)
Alan Lee

Abstract: We survey methods for generating synthetic data sets without making use of unit-record data. The methods we describe are based on creating data sets which match publicly available marginal tables. We describe a set of R functions which implement the methods under study, and apply the methods to data from the 2001 Census of Population and Dwellings.

Keywords: Contingency tables, marginal tables, log-linear models, maximum likelihood, iterated proportional fitting, integer programming.

Disclaimer

Official Statistics Research, through Statistics New Zealand, commissioned these reports. The opinions, findings, recommendations and conclusions expressed in the reports are those of the author(s), do not necessarily represent Statistics New Zealand and should not be reported as those of Statistics New Zealand. The department takes no responsibility for any omissions or errors in the information contained in these reports.