The future of microdata

Professor Denise Lievesley, Special Advisor, United Nations Economic Commission for Africa

Denise will talk about the importance of preserving and providing access to well documented microdata from population censuses, other censuses and key surveys in order to support evidence-based decision and policy making within countries. She will look at this issue from a global perspective, focussing particularly on Africa. She will discuss some of the barriers which have prevented the development of access to microdata and will highlight recent developments (in relation to the establishment of an Association of African data archivists, the re-writing of national statistical legislation and the development of national statistical strategies) designed to improve the current situation whereby census data are still under-utilised within so many countries. Denise will also discuss the empowerment of the user community which can result from access to census data.

International Microdata Revolution

Managing Microdata Access and Statistical Confidentiality: A review of the IPUMS-International project
Dennis Trewin, Statistical Consultant

The 2003 Conference of European Statisticians was the first occasion the heads of National Statistical Offices collectively considered the question of the provision of microdata to support research. Importantly they agreed that supporting research is an important activity for National Statistical Offices (NSOs) and that most NSOs could do more to satisfy these needs. Doing more included providing improved access to microdata. They commissioned work which led to the publication of Guidelines on Confidentiality and Microdata Access (published on the web site of the Conference of European Statisticians). The Guidelines were later adopted by the UN Statistical Commission for international use. I was the leader of the Task Force which undertook the work and author of the Guidelines.

In this light I was contracted by the managers of IPUMS-International to review its conformance with international best practice. IPUMS-International is one of the world’s largest repository of microdata. The talk will highlight the main findings of the review. It will also attempt to draw out the main learnings for other repositories of microdata.

Intergenerational Coresidence and Economic Development: New Evidence from the International Integrated Public Use Microdata Series
Steven Ruggles and Misty Heggeness, Minnesota Population Center

We use newly-available census microdata from IPUMS-International to assess trends in intergenerational coresidence in 15 developing countries. Contrary to expectations, we find no general decline in intergenerational coresidence over the past several decades. There have been, however, significant changes in the configuration of intergenerational coresidence. Families in which the older generation is household head—a configuration consistent with traditional patriarchal forms in which the older generation retains authority—are becoming more common in most of the countries. Intergenerational families headed by the younger generation—the configuration one would expect if intergenerational coresidence were motivated by a need for old-age support—are on the decline in most of the countries. Multivariate analysis reveals that intergenerational families headed by the older generation are positively associated with measures of economic development. These findings are at variance with widely-accepted social theory. We hypothesize that housing shortage, economic stress in the younger generation, and old-age pensions may contribute to the change. More broadly, in some developing countries rising incomes may have allowed more people to achieve their preferred family structure of intergenerational coresidence following traditional family forms.

Lessons learned from cross-national research on marital homogamy
Albert Esteve Palós and Luís Ángel López, Center for Demographic Studies, University Autonomous of Barcelona

Full abstract (Word)

This paper discusses the advantages and drawbacks of carrying out cross-national research on marital homogamy using census microdata. The integration of census microdata at the international level by the IPUMS projects (Minnesota Population Center) has opened the door for cross-national research on a broad range of topics, including assortative mating. Census microdata are an invaluable resource for social science research. Other sources—such as demographic and labor force surveys—often offer greater subject coverage and detail than do census data, but no alternate source offers comparable sample density, chronological depth, and geographic coverage.

The Census-project: Evaluating the IECM base and enhancing users' access
Sabine Springer, CENSUS-Project, French National Institute for Demographic Studies

Full abstract (Word)

The overall objective of the CENSUS project consists in the evaluation of the coverage and quality of the Integrated European Census Microdata base through demographic methods. This database, which is still under construction, has been built in cooperation of the Minnesota Population Center and the Centre d’Estudis Demogràfics of Barcelona. The countries already available and which will be considered in the CENSUS project are Belarus, France, Greece, Romania, Spain, Hungary and Portugal, a total of 7 countries and 22 censuses. The final outcome will be guidelines for good practices for potential users of the database prepared by the three participating research groups: Research Unit for International Comparability (INED), the European Demographic Observatory and the Institute for demographic studies of the University of Bordeaux. The project is financed by the National Research Agency of France (ANR).

Health inequalities using Belgian census and linked mortality data
Patrick Deboosere, Department of Social Research, Vrije Universiteit Brussel

Administrative data is not designed for research but rather a by-product of “administrative bookkeeping”. Computerization has played a major role in the explosion of the research potential of administrative data for demography. The centralization and computerization of the Belgian municipal registers in the National Population Register became fully operative in 1988 and boosted a new research field on the analysis of administrative data. The linkage of core socio-demographic data from the census with event data of the national population register has been applied extensively in several Scandinavian countries proving how this kind of research design contributes to evidence based scientific progress in demography, epidemiology and sociology. As the cost of linking data is relatively small and resulting databases are highly informative, the re-use of administrative data for statistical purposes for research and policy has become increasingly popular. Monitoring evolutions is becoming crucial in our complex modern societies in order to develop efficient policies. Population health is no exception in this regard. The inclusion of a small set of health questions in the Belgian census and the linkage with mortality data demonstrates the potential of administrative data combined with simple single-item questions for the monitoring health inequalities.

Comparing results from the England and Wales, Scotland and Northern Ireland Longitudinal Studies
Harriet Young and Emily Grundy, London School of Hygiene and Tropical Medicine, Paul Boyle, St Andrews University and Dermot O’Reilly, Queens University Belfast

There are now three separate record linkage studies which include census and vital registration data covering England and Wales (the ONS LS), Scotland (Scottish LS) and Northern Ireland (NILS), managed in three separate statistical offices. The aim of this project is to develop methods and procedures for carrying out parallel and if possible combined analyses of these three studies. To do this, we use an exemplar research project to analyse the association between self reported health and long-term illness in the 2001 Census and subsequent mortality in all three studies. Our research objectives are first to analyse socio-economic and socio-demographic variations in reporting long limiting long-term illness and self rated health in 2001 within England & Wales, Scotland and Northern Ireland; and second to analyse associations between reporting of health status in 2001 and subsequent mortality (taking account of socio-economic and socio-demographic factors). This is an important issue as there are some suggestions of differing associations in constituent parts of the UK. Here, we report on our progress in carrying out parallel analyses, using the research project as an example to illustrate this.

Socio-demographic determinants of entry into and exit from long-term institutional care – a linked register based follow-up study of older Finns
Pekka Martikainen, University of Helsinki

Full abstract (Word)

For most elderly people long-term institutional care is often the last resort when physical and mental functioning fails. Furthermore, publicly funded long-term care takes up, on average, about one percent of GDP in OECD countries, and this may more than double by 2050. Thus, in terms of considering future policy choices and estimating long-term care costs, it is important to understand the determinants of long-term care. The aim of this study is to assess the effects of socio-demographic determinants on entry, exit and length of stay from long-term institutional care.

Migration

Using IPUMS data from the 1999 Kenya Census to explore internal migration
David Owen, University of Warwick

Full abstract (Word)

Though it has been one of the most politically stable countries in Africa since achieving independence in the 1960s, Kenya faces substantial challenges in achieving the economic development necessary to lift its rapidly growing population out of poverty. There are substantial geographical differentials in economic development and levels of income and poverty within the country.

While sample labour force, health and poverty surveys are regularly conducted in Kenya, the most recent comprehensive data on the population is still the 1999 Population and Housing Census (preparations are currently being made for the 2009 Census). A five per cent microdata sample from the Census is made available via the IPUMS data archive at Minnesota.

This paper makes use of the IPUMS microdata for Kenya to explore patterns of internal migration within the country and the associations between migration and poverty differentials. The dataset provides information on demographic and economic characteristics of individuals within households, and also records an individual’s district of birth, district one year before the Census and length of residence at their current address.

A Gender Approach to Cultural Norms and Labour Migration: The Case of Migration from Romania
Raluca Prelipceanu, The Sorbonne Centre for Economics, University of Paris

Full abstract (Word)

Our paper analyses the determinants of the Romanian labour migration during the transition period. Our approach is comparative taking into account the differences between the migration decision of men and that of women. Our main research questions are “Do migration determinants differ according to gender?” and “Do local norms influence the propensity to migrate of women and that of men?”. In order to answer these questions, we build first of all a collective household model of the decision to migrate. We consider the household as being made of two members: the husband and the wife. The wife works on the labour market, but is also involved in the production of a domestic public good, in this case children, whereas the husband works only on the labour market. We consider the market for childcare to be incomplete so the household cannot employ another person to take care of the children. However, there is imperfect substituability between family members inside the household. As children are part of the utility function of both the husband and the wife, a decrease in the utility function of the children would also induce a decrease in their parents’ utility functions. Due to differences in the domestic labour supply, the migration cost in the case of women exceeds the migration cost of men. We consider this to be one of the main explanations for the low mobility of women. In addition, the migration decision is also influenced by the sharing rule which, in turn depends on the income of the family members and on distribution factors like gender and cultural norms.

Migration in National Surveys and Censuses
Adriana Castaldo and Gunjan Sondhi, University of Sussex

Full abstract (Word)

In recent years, as part of the overall increasing importance that the topic of migration - from and within developing countries - acquired globally, it has been amply acknowledged that improving the availability and the quality of data on migration flows and the characteristics of worldwide migrants is key to the achievement of an informed and sound design of migration policies. This applies both in the areas of the migration of adult individuals, as well as children. In an attempt to fill in the existing gap on the availability and use of quantitative data on adult and child migration, we have developed a unique catalogue of existing national-level household surveys and censuses in developing countries. This collection of household surveys and censuses contains information on and access to data that can be used to analyse child migration and forms an integral part of the catalogue “Migration in National Surveys” (MiNS).

This paper describes the MiNS catalogue and uses part of the data collected in the catalogue as the main tool to investigate how the theme of migration and specifically child migration is treated in some of the censuses; how this data has been and could be used in the study of child migration; and to address issues of data quality and access to data.

Using the 2001 Census Individual CAMS to explore the relationship between migration and longer distance commuting in rural England
Tony Champion, University of Newcastle

Full abstract (Word)

For some time there have been concerns about the effect of in-migration on the life of rural communities and on the sustainability of rural settlement more generally. Particular issues have been raised by the movement of people out from cities who continue to work there and therefore have long work journeys. Yet the evidence base on the links between migration and commuting has traditionally been weak, partly due to the dearth of large-sample datasets covering both of these behaviours in any detail. Even with the Population Census, it is only with the advent of the SARs that – short of obtaining commissioned tables – it has been possible to crosstabulate people on the basis of migrant status and journey-to-work pattern.

The research results reported in this paper take advantage of the richness of the data available in the Individual version of the Controlled Access Microdata Sample (CAMS). With its 3.125% sample of England’s residents, it contains derived variables on distance to work and on distance between current usual residence and that of one year ago. It also gives the local authority of both current and previous address, allowing a classification of migrants by the types of place moved from and to. Along with the detailed breakdown of all the personal characteristics collected by the Census, this makes it a powerful source for answering the study’s central question about how much the odds of being a longer distance commuter are affected by being a recent migrant and by the type of move made.

Health and Employment

South Asian ethno-religious groups and part time employment pattern in England and Wales: Results from the 2001 individual SARs
Reza Afkhami, University of Manchester

Full abstract (Word)

In recent years it has become clear that some minorities face larger penalties than others in the labour market (Craig, et al. 2005; Dale, et al 2006, 2002; Fieldhouse and Gould, 1998). It is also known that the nature and perhaps the extent of discrimination faced by ethnic minorities in Britain vary between different ethnic minority groups mainly Muslim south Asian, Modood (1997). A recent work by Simpson et al (2006) indicates a further research needed to investigate high rates of part-time working among some populations including Pakistani and Bangladeshi men. This is particularly important as Part-time working may imply lower incomes, poorer working conditions and barriers to full-time work.

This paper analyzed part time working among south Asian population in England and Wales. We tested whether south Asian ethno-religious composition can explain the different patterns of part time working in England and Wales.

Measuring Social Progress: Labour market position of 1st and 2nd generation minority ethnic groups in Britain (1991-2001)
Yaojun Li, University of Manchester

Full abstract (Word)

Ever since the 1991 Census, particularly the advent of the samples of anonymised records (SARs) at the individual and household levels with the ethnic data available, there has been a huge interest in the economic fates of the minority ethnic groups in Britain. Many studies were conducted as can be seen in the collection by Karn (1997, for a summary of the key findings and research potentials of the SARs, see Li 2004). This interest has also influenced the collection and analysis of data, particularly social surveys, by other government agencies and academic community. Indeed, there has been much more academic and policy interest in ethnicity than in the traditional mainstream sociological issues such as mobility and gender inequality. Yet, it is also the case that little systematic research is available on the social progress, or lack of it, in terms of the labour market position by first and second generation minority groups in Britain across the decade using standardised variables and methods.

This paper will use the 1991 and the 2001 SARs to address the issue. We shall use standardised variables at the individual level on ethno-generational status, labour market position in terms of employment and occupational (class) status, and a range of other personal attributes and human capital variables such as age, education and marital status as available in the SARs. We shall also construct and add new variables on the socio-economic development (GDP/GVA) and the ethnic diversity at the regional level at both time points to serve as contextual factors. The two sets of variables will enable us to conduct a systematic research on trends of ethnic penalty in British society over the period covered.

The Geography of Unpaid Caring in the UK
Paul Norman, University of Leeds and Kingsley Purdam, University of Manchester

A person is a provider of unpaid care if they give any help or support to family members, friends, neighbours or others because of long term physical or mental health or disability, or problems related to old age. It is estimated that there are over 6 million unpaid carers in the UK of whom 1.7 million provide over 20 hours of care per week. The number of ‘informal’ carers may need to double within a generation due to the ageing of the UK population. In this paper we examine geographic and social variations in the amount of unpaid caring across the UK. Using the Samples of Anonymised Records from the 2001 Census we map sub-national geographic variations in the amount of caring in the UK. We consider variations in caring at the individual level using a specially devised UK index of caring. The findings have important implications for our understanding of the dynamics of caring and for service providers at a national and local level.

Uses of census microdata for monitoring human resources for health in low and middle income countries
Neeru Gupta, World Health Organization, Switzerland

Full abstract (PDF)

Human resources are a strategic capital in any organization, but especially so in health and other service organizations which are highly dependent on their workforce. Despite the undoubted importance of monitoring human resources for health (HRH) and the impacts on health systems
performance and population health outcomes, in many countries the empirical evidence to support policy formulation is often fragmented and incomplete. Many sources that can potentially produce statistics on the health workforce remain underused. In particular, the potential of population censuses for HRH monitoring has generally not been met, especially in developing countries where access to microdata has generally been limited and comparisons across countries or time periods are difficult because of inconsistencies in both data and documentation.

The World Health Organization's Department of Human Resources for Health works to strengthen the global evidence base on HRH through the building of capacity for analysis and use of data from standard statistical sources, including census sources. This paper has three main components. First, we discuss the relative strengths and limitations of census data for health workforce analysis. Second, we present findings from recent research on HRH stock and distribution using microdata from the Integrated Public Use Microdata Series (IPUMS)-International databank for the 2000 round of censuses for selected low and middle income countries. Lastly, we explore opportunities for strengthening global health workforce analyses in the upcoming 2010 round of censuses.

International Migration

Status, class and immigration in the 1901 Census: results of applying 19th Century and 20th Century classifications to occupational data on immigrants and the native-born in the 1901 Census of Canada
Charles Jones, University of Toronto

Full abstract (Word)

Sociologists and Historians agree that occupation is an important indicator of an adult’s place in the social stratification system. However they have quite different views about whether stratification is most usefully defined as a continuous socioeconomic gradient or as a set of discrete social classes. Blau and Duncan’s seminal work, The American Occupational Structure uses both categorical and continuous approaches but more recent North American empirical research has mostly followed Duncan’s approach in seeking and using “a socio-economic index for all occupations” (Duncan’s SEI). This is seen most clearly in the IPUMS project where Ruggles and other members of the Minnesota group have used the 1950 US Census classification and SEI in order to code occupation data in all US Censuses from 1850 to the present. More recent non-US censuses and surveys have used the International Standard Classification of Occupations and an international socioeconomic index scale. Taking another direction there are now several 19th Century Census data files from countries bordering the North Atlantic where the occupational information has been coded according to the North Atlantic Population Project’s modifications of the more social-class oriented Historical Standard Classification of Occupations (HISCO): a classification that is more appropriate to agrarian societies with pre-modern social relations of production and large numbers of domestic service workers.

Family Formation of ‘Foreign Brides’ in Japan and Taiwan: A Comparative Analysis of Two Types of Censuses
Hiroshi Kojima, Waseda University, Tokyo

Full abstract (Word)

A Rapid increase in international marriages and the social integration of foreign female spouses (wives) and their children became a social issue in Taiwan. Consequently, the Taiwan Ministry of the Interior conducted the Census on Living Conditions of Foreign and Mainland Spouses in 2003 with all the foreign and Mainland spouses of Taiwanese nationals. In Japan the integration of foreign wives and their children became an issue earlier than in Taiwan, but the support measures were largely left to local governments and NGOs and the information is largely based on small-scale surveys. In fact, the Population Census is the only source of nationwide information on couples in international marriage in Japan. The purpose of this study is to clarify the determinants of family formation behaviors of foreign wives in international marriage in Japan and Taiwan to supplement insufficient information and to derive implications for Japanese official statistics, drawing on microdata from the 2000 Population Census of Japan and the 2003 Taiwanese census of foreign spouses.

Gender Ratios in Global Migrations, 1850-2000
Johanna Leinonen, University of Minnesota

Full abstract (PDF)

Scholars in many disciplines have observed that men historically far outnumbered women among international movers. Two of geographer E.G. Ravenstein's latenineteenth century "laws of migration" asserted that (1) short-distance migrants generally outnumbered longer distance ones, and that (2) within-country moves were usually dominated by women and between-country moves were dominated by men. Although many geographers have critiqued and extended Ravenstein's work, these two oftenrepeated laws have rarely been challenged since they were written in the late 1800s. Textbooks in demography and world history presented these theories as conventional
wisdom for much of the twentieth century (e.g., Peterson 1969: 264; United Nations 1979: 4; Manning 2005: 11).

In recent years, researchers at the U.S. Department of Labor and elsewhere have pointed toward a "remarkable shift" in migrant gender ratios, from women constituting less than one-third of all U.S.-bound migrants in 1900 to almost one-half in the 1970s (Houston et al. 1984: quote on 913, Simon and Brettell 1986). The United Nations' 2006 State of World Population Report advised policy-makers to take note of the fact that "today, women constitute almost half of all international migrants worldwide" (United Nations 2006). The U.N. report outlines a broad range of policy issues—such as social services, labor market policies, and migration regulations in sending and receiving countries—that are affected by this apparent shift in the gender composition of migration
streams.

Combining data

Combining 2001 UK Census data in a multilevel model
Mark Tranmer, University of Manchester

I explain how the research value of Census data is maximised by combining census aggregate and micro-data in a multilevel modelling framework, and illustrate this approach given the data availability from the 2001 UK census. I show how multilevel model parameters may be estimated, even though the available Census data are not of standard multilevel form. Moreover, I explain the important role of micro-data in this estimation process. Finally, I provide some empirical results based on an analysis of the 2001 census data, and simulations based on the 2001 data structure. These results suggest that there it is highly beneficial to combine aggregate and micro data in this situation, rather than working with a single source such as the census aggregate data.

These methods will benefit Census data users as well as having more general impacts in social statistics and human geography. The methodological developments are also relevant to many other situations where multiple sources of data for the same population are combined, with implications for design, analysis, confidentiality protection and data-release.

Using Targeted Perturbation of Microdata to Protect Against Intelligent Linkage
Mark Elliot, University of Manchester

Full paper (Word)

This paper describes linkage experiments using the 2001 individual Sample of Anonymised Records (SARs) from the UK census to the microdata output from the UK Labour Force Survey (LFS) for spring 2001.

The objective of the study was to assess the impact of the statistical disclosure control methods on the used on the 2001 SARs on the ability to link an external dataset and SARs. The Labour Force Survey was selected as the external file because (i) it was of sufficient size to produce a large enough overlap with the SARs and was collected around the census date.

The project follows the tradition of other such studies with official data; e.g. Muller, W; Blien, U.; and Wirth, H. (1992), Elliot and Dale (1998). However, the study here elaborates on that earlier work by examining the impact of a targeted disclosure control technique on the ability of an intruder to attack a dataset by focusing on the high risk records.

Family formation

The Mixed-Ethnic Unions in England and Wales in the 1990s

Zhiqiang Feng, University of St Andrews

Full abstract (Word)

Although developed societies are becoming increasingly ethnically diverse, relatively little research has been conducted on mixed-ethnic unions (married or cohabiting). Those studies which have been undertaken demonstrate that mixed-ethnic unions account for a small fraction of total couples, but they are growing in number. In England and Wales, the number of mixed-ethnic unions increased by 65 percent between 1991 and 2001. Previous studies in Britain tended to focus on a general picture of the number and trends of mixed-ethnic unions (Berrington, 1996, Coleman, 2004). One exception is the study by Muttarak (2004) which uses Labour Force Survey (LFS) to analyse the factors contributing the formation of mixed-ethnic unions. Our study uses data from 1991 and 2001 Household Samples of Anonymised Records (Household SARs) which are based on a 1 percent sample of the total households in England and Wales. The Household SARs provide information on both partners in the family, which allows us to identify mixed-ethnic unions and the large sample size allows us to investigate mixed-ethnic unions by individual ethnic group. In the study we document the growth of mixed-ethnic unions by ethnic group in the 1990s. We also examine whether there are differences between mixed- and co-ethnic unions in terms of demographic and socio-economic characteristics, focusing particularly on age, generation, education, and social class. Furthermore, we use log-linear models to investigate whether the propensity of out-partnering for different ethnic groups has changed between 1991 and 2001.

Studying links between disability and family formation: Evidence from the SARS and other data sources
Stephen McKay, University of Birmingham

Full abstract (Word)

There is mixed research evidence on the links between family formation and disability. Some studies of specific illnesses and health conditions have found strong links, with disability-related conditions increasing the risk of family breakdown, whilst other studies have found no link. Few such studies have been conducted in the UK. The other main limitations of UK-based research has been relatively small sample sizes, concerns about the appropriate measurement of disability, and especially over-reliance on a medicalised model of disability, and the lack of an overall structure in which to conceptualise family transitions.

The study examined:

1. Family breakdown – that is, splitting up among both married and cohabiting couples.
2. The rate of single people becoming couples (‘partnering’).
3. Effects on family size, through any changes in the timing of births and whether they happen at all.

Poverty Mapping

Micro-level and Macro-level Effects of Family Poverty in the Texas Borderland and the Lower Mississippi Delta: United States, 2006
Dudley L. Poston, Texas A&M University

Full abstract (Word)

The Texas Borderland and the Lower Mississippi Delta are the two poorest regions in the United States (see Figure 1 below). Each is characterized by high concentrations of racial and ethnic minorities, Mexican Americans in the Borderland and African Americans in the Delta. The high and persistent poverty suffered by the residents of these regions has much to do with the disadvantaged socioeconomic position of racial and ethnic minorities in the United States.
This paper uses microdata from the 2006 American Community Survey for families residing in the Borderland and Delta. We first estimate the log odds that a family is poor. We use three different definitions of family poverty: 1) whether or not the family has annual money income that is 50% or less of the official poverty threshold (“in deep poverty”), 2) whether or not the family has annual income that is 100% or less of the official poverty threshold (“in poverty”); and 3) whether or not the family has annual money income that is 150% or less of the official poverty threshold (“near poverty/in poverty”).

Census microdata and poverty mapping – a validation study using Brazilian data
Pedro Luis do Nascimento Silva, University of Southampton

Full abstract (Word)

‘Poverty mapping’ has attracted widespread attention in recent years. This is an approach to the estimation of poverty indicators at small area level developed in response to the need for targeting public policies to alleviate poverty. An approach which is currently advocated by the World Bank and has been used in several countries is based on combining high quality information on income (or expenditure) collected using sample surveys designed to estimate for national or broad-level geographies, with detailed information (microdata) provided by a recent census of population (Elbers, Lanjouw and Lanjouw, 2002 – abbreviated ELL). The idea is to fit a model to explain income (or expenditure) using the sample survey data, with predictors which are also measured on the census. Then the model is used to perform ‘mass random imputation’, namely imputing income (or expenditure) for each census record (household). The imputed values are used to calculate the target poverty indicators at the level of the small areas of interest. Multiple replicates of the imputed census file are used to average the poverty indicators and to obtain corresponding standard error estimates.

A Technique for Small Geographical Scale Poverty Analyses: The use of Census Microdata in the Case of Liberia, Costa Rica
Carlos de la Espriella, Lund University, Sweden

Full abstract (Word)

Poverty analyses are instrumental for a pro-poor urban planning practice, but they seem weak in addressing the causes of poverty at city level in developing countries, because they are seldom articulated in a strategic set of pro-poor local actions, and they do not reach small geographical scales at city level where actions should be implemented.

In an attempt to bridge this methodological gap, a technique for undertaking poverty analyses at city level was tested in the analysis of the implementation of Costa Rica’s social housing policy in the medium-sized city of Liberia. The analysis used 2000 National Population Census data of the city of Liberia, desegregated at small geographical units .

The Research Data Centre Program: from a seed to a forest. A fundamental element of the social research infrastructure in Canada

Gustave Goldmann, Statistics Canada

Informed decision making on social issues requires current, comprehensive and very well-targeted research. Societies face two primary challenges in order to respond to this need for timely information – access to relevant data and a corps of qualified researchers to conduct the analyses. As part of a response to the challenges that confront Canadian policy research, a network of Research Data Centres was formally launched in December 2000. There are currently 25 Research Data Centres and Branches located throughout the country, so researchers are not obliged to travel to Ottawa to access Statistics Canada data. At the same time, the centres are administered in accordance with all the confidentiality rules required under the Statistics Act. The Research Data Centres meet, in a single location, both the need to facilitate access to detailed micro-data for crucial social research and the need to protect the confidentiality and security of Canadians’ information. The research conducted in the centres contributes significantly to the public debate on key issues that are of concern to Canadian society. The network expands the collaboration between Statistics Canada, the Social Sciences and Humanities Research Council, The Canadian Institutes for Health Research, universities and academic researchers. It is also instrumental in training a next generation of Canadian quantitative social scientists. The Canadian RDC Network serves as a model for achieving a balance between providing open access for research and respecting the confidentiality of official data.

Improving access to Census Microdata - the Australian experience

Jenny Telford, Australian Bureau of Statistics

With the increasing pressure to release more detailed data to clients, protecting the confidentiality of respondents has never been more important than now.

The Australian Bureau of Statistics, like statistical agencies around the world, is committed to protecting the privacy of our respondents and preventing any form of unit level identification. This presentation will focus on the methods being developed by the ABS to improve the way clients access and interrogate microdata while balancing the risk of disclosure with procedural and technical protections. We will outline the advances made through the 2006 Census output program with the development of TableBuilder and provide an overview of the Remote Access Data Laboratory.

Microdata access and confidentiality issues in Germany

Heike Wirth, ZUMA, Germany

Full abstract (Word)

The focus of this presentation is on the possibilities of microdata access in Germany, which might be slightly different to other countries. Along the lines “chi va piano va sano” the microdata access has improved slowly but continuously in Germany over the last two decades. The most important improvements are reflected in (1) an increase of producers of official microdata making their data available for research purposes; (2) a standardization of data access; (3) different types of data access. Microdata can be used in terms of (a) scientific use files (off site access); (b) on site access; (c) remote access. Regarding the flexibility of data analysis (in terms of e.g. time, costs and keeping track of research ideas) scientific use files are the type of microdata access preferred by most researchers. Scientific use files are microdata which are anonymized according to the principle of ‘reasonable anonymity’ (faktische Anonymität) and are released for academic research only. ‘Reasonable anonymity’ means that the reidentification of respondents is not impossible (‘absolute anonymity’) but comes along with a disproportionate amount of time and costs.

Census futures

Routes to access: what affects the decision that statistical offices make on research access
Paul Jackson, Jane Longhurst and Judy Hawkins, Office for National Statistics, UK

The primary debate is over. With suitable safeguards, it is appropriate for a statistics office to provide research access to confidential data. For ONS, this is confirmed by features of the Statistics and Registration Service Act. This presentation will look at the real-world factors that have determined the routes to confidential ONS data currently on offer. It will also look at the pinch points that will shape future routes to access, and how the research community can help with addressing them.

The Danish system for researchers access to microdata

Ivan Thaulow, Danmarks Statistik

Full abstract (Word)

The access of researchers to Danish microdata as well as the quality of the available administrative data in Denmark has improved significantly in the recent years. Through the Research Services at Statistic Denmark researchers have now access to a large number of administrative registers in many fields of research such as labour market research, sociology, epidemiology and business economics.

In Denmark researchers are only given access to de-identified microdata. However, the strength of the Danish system is that we use almost universal identification keys (such as “person numbers” and “central business register numbers”) in most of the administrative registers. Thus it is possible for Statistic Denmark to correlate the aggregated data before data is de-identified and made readable to the researcher. Data can be correlated and aggregated both within a specific year and longitudinally across several years. This creates unique possibilities to combine knowledge over time from different administrative registers in order to be used in social and medical research.

The UK Census: future directions
Peter Fullerton, Office for National Statistics

The decennial UK Census is one of the most important statistical data sources in the UK. Apart from its wide use in government, academia and the commercial sector it is the key data source to measure the number of people living in the UK, the households they live in, and their key individual characteristics.

Although it is recognised the Census is an important data source, there have been recent developments in the population which make it more difficult for a Census to provide up-to-date measures. Because of the increase in migration (both internal and external) a decennial measurement of the population is seen as insufficient and it poses some serious challenges to government in the way it sets policies and allocates local funding for services. This gap is increasingly being recognised by a number of independent observers and has been the subject of criticism by the Parliamentary Treasury Committee.

In responding to this criticism the Office for National Statistics recognises the need for improvements in the way we measure the population. In looking at alternative approaches it is clear that there are many data sources available which, if used in combination, could provide the potential to improve measurement of a mobile population at small area levels. The presentation will set out some of the key issues and will consider advantages and disadvantages of alternative approaches to count the population and its characteristics.

French census microdata: changes ahead

Jean-Francois Royer, Institut National de la Statistique et des Études Économiques (Insee), France

Full abstract (Word)

The last traditional census in France took place in 1999. The micro data sets available for researchers from that operation have been limited: first by the confidentiality rules of the French statistical law (“loi de 1951”) which does not contain exceptions for research as far as personal data are concerned; second, by the supplementary rules adopted on the recommendation of the “National commission for data processing and liberties (CNIL)” in order to prevent the risks of “profiling”. These rules define some information as “sensitive” (nationality, country of origin, date of first entrance to France) and put thresholds on the size of the geographical units that can be described by micro data files; these thresholds vary according to the nature of variables contained (sensitive or not, detailed or not) and according to the sampling rate. For example a 1/20th dataset has been put in public use, with almost all the variables (but not the sensitive ones, nor the census district); a 1/4 dataset has been created for research, with all the variables except detailed geographical ones (under the “region” code), etc. Special provision has been made for local authorities: they can receive more detailed information than ordinary users. For researchers who need more information than the public one, the only way to process census data is to ask INSEE for customized tables (this is a paying service, through a dedicated laboratory “Centre Quetelet”). Special mention can be made of two derived files: “SAPHIR” and “EDP”. The first (SAPHIR) is a collection of micro-data of several French censuses (from 1962 to 1999) with coordinated codes. The second (EDP : “échantillon démographique permanent”) is a 1/100 sample of census and register data collected and matched for the same people (panel). These files have been kept inside INSEE for confidentiality reasons, and only researchers working with INSEE have been able to use them.

The Census Public Use Microdata File (PUMF) at Statistics Canada
Sri Kanagarajah, Census Operations Division, Statistics Canada

Full abstract (Word)

From 1971 to 2001, Statistics Canada have been producing three single files with no relation between the files: Individuals; Families; Household and dwelling. After consultations with users (academic, private sector and federal department), they promoted a new approach for 2006 PUMF. They would like to create one single file for individuals (size: 2.7% of the population) and one hierarchical file (size: 1 % of the population) with individuals, family and household and a link between the three universes. This allows users to link and study relationships between individuals and their families or households as well as allows international comparison.