Seminar Long Summary Report

Workshop: Rethinking the Estimation and Projection of Urban and City Populations New York City, 9-10 January 2006

Overview
This meeting brought together demographers, experts in remote-sensing applications and geographic information systems, urban planners, and representatives of relevant United Nations agencies, providing them with an opportunity to engage in discussions on new sources of data and emerging methods in urban population estimation and forecasting, with a decided emphasis on methods that make use of spatially disaggregated data. This highly multidisciplinary workshop demonstrated that in addition to its considerable scholarly appeal, the integration of diverse data and methods holds great promise for urban policy and planning in poor countries.

The Workshop was sponsored by the IUSSP under the auspices of its Scientific Panel on Urbanisation. That panel includes the following IUSSP members: Mark Montgomery (chair), Deborah Balk, Eduardo Moreno, and Thomas Buettner. The Workshop was co-funded by the US National Science Foundation Advance Project to the Earth Institute of Columbia University and the United Nations Population Division. The Workshop was held in New York City, at Low Library of Columbia University.

Fifty individuals from North and South America, Europe, Africa and Asia participated in the Workshop. They included demographers, sociologists, economists, urban planners, geographers, and geophysicists (with an expertise in remote sensing technologies). Although the majority of attendees were from North America, they represented international organizations as well as universities and other agencies in the US. All of the US-based researchers were actively engaged in urban work that was either international in scope or had a strong involvement in at least one developing country. About one-third of the participants were women. Slightly more than one-third of the participants represented junior scholars, about half of whom were women.

The 2-day meeting was full of exciting presentations and stimulating dialogue among this highly diverse and interdisciplinary group of participants. The participants took seriously the need to reach beyond their disciplines to address the core issues: how best to merge information from different sources to gain a fuller understanding of urban population and growth dynamics.

Summary of Presentations
The workshop was divided into four main areas of inquiry: (1) Detecting Urban Footprints; (2) Measuring the Components of City Growth; (3) New Forecasting Methods; and (4) Institutional Investments in Data. A set of questions to guide the panelists and the discussion was circulated before the workshop, and is available at: http://www.iussp.org/Activities/wgc-urb/urb-questions06.php.

The Workshop began with three stage-setting presentations. Dr. Hania Zlotnik, Director of the Population Division of the United Nations and former Vice-President of the IUSSP, reviewed the main intellectual and policy contributions as well as the hurdles to understanding urbanization over the past 30 years. Dr. Uwe Deichmann, of the World Bank, spoke on the contributions of spatial social science. Dr. Gordon McGranahan, of the Institute for the Environment and Development, spoke on the value of city and urban projections for development.

Recent advances in new methods and data collection are promising. These advances were reviewed in three presentations led by the co-organizers, in a session “Where Do We Stand?” which was chaired by Dr. Paul Cheung, Director of the Statistics Division of the United Nations. Professor Tony Champion of the University of Newcastle began with a review of the lessons from the preceding IUSSP Working Group on Urbanization, which led to a call for geographic data to be made available at the building-block level to allow for greater comparisons and flexibility across places and institutions. Dr. Thomas Buettner, of the United Nations’ Population Division, then reviewed the current state of their Cities Database. Dr. Deborah Balk, of Columbia University, showcased a recent global effort (the Global Rural Urban Mapping Project, GRUMP) at detecting the spatial extents of urban areas. Dr. Mark Montgomery, of the Population Council, then illustrated new ways in which the Cities Database and GRUMP could be linked with demographic surveys in models of city growth. Dr. George Martine facilitated this discussion.

Concluding the first morning session, a lunch-time presentation from an applied perspective, “Urbanization and populations at risk: Spatial data for malaria burden estimation” was given by Dr. Andy Tatem of Oxford University.

Detecting Urban Footprints
The leading question for the panel of urban footprints was: What can be learned about urban areas from remotely-sensed (i.e. earth observing satellite) data? For the purpose of the detecting trends in urbanization, three areas are key: (1) detection of urban footprints; (2) detection of changes in spatial extents over time; and (3) an ability to detect intra-urban variation. The panelists presented three differing perspectives on these issues: Dr. Chris Small, of Columbia University, discussed the contributions of optical imagery at moderate and high resolutions. Dr. Chris Elvidge, of NOAA, also presented on optical imagery, but discussed the global “night-time lights” data set. Finally, a presentation was given by Dr. Son Ngheim and Dr. Ernesto Rodriguez of the Jet Propulsion Laboratory on the potential uses of radar data. Dr. Steve Sheppard led the discussion, presenting insights from his study of detecting urban change in an economic context for a large sample of cities using moderate-resolution satellite data.

There was a general agreement that remote-sensing data has considerable promise for evaluating change, and even at detecting intra-urban features. But, since remote sensing technologies and data classification are firmly rooted in the natural sciences, the generated data are not highly compatible with social science data formats, and substantial training is required to use them. It was evident that more interdisciplinary engagement—and perhaps some small-scale pilot projects—would be helpful in furthering social science applications of urbanization.

Measuring the Components of City Growth
This panel, on measuring the components of urban growth, was organized around the following leading questions: Is it advisable for estimates and projections of city population growth to make use of data on urban fertility and mortality rates? Since urban fertility and mortality survey data is rarely representative of individual cities, how should we take advantage of data that represent broader geographic areas (e.g., regions and sub-regions within countries)? What is the role of national demographic surveys such as the DHS and MICS in providing fertility and mortality data, as against censuses and other local sources of information? What would be the ideal frequency of observation? Can global estimates be assembled using a mix-and-match approach (surveys, censuses, vital registration)? Are we yet in a position to include components of urbanization that have been difficult to formalize, such as migration? Can we improve our estimation of the urban poor in particular? To what extent does spatial information aid in the development of new methods?

Several perspectives were given in this panel. Dr. Renato Assunca of Universidade Federal de Minas Gerais illustrated the use of Bayesian statistical methods to generate small area estimates of urban fertility and mortality schedules in Brazil. Dr. James Raymer of the University of Southampton showed recent advances in indirect estimation of migration and the scope for multi-regional projections. Two regional perspectives were also offered: Professor Kam Wing Chan of the University of Washington described some of the regional challenges in modeling growth with a particular emphasis on China, and Dr. S. Chandrasekhar of Indira Gandhi Institute of Development Research described challenges in measuring the nature and extent of slum populations in India.

This session confirmed that recent advances in estimating the components of growth have the potential to replace traditional methods of estimating the components of demographic change. In particular, surveys can and should be used with new methods, as should vital registration data where these are available; migration data are increasingly available (though still seldom estimated at the city-specific level), and determining who and where are the urban poor—if not straightforward—can be achieved with some effort. The case studies of China and India also demonstrated that an understanding of the urban dynamics of countries may be an especially complex undertaking that requires considerable country-specific expertise as well as input from international researchers and agencies.

New Forecasting Methods
The panel of new forecasting methods addressed the following leading questions:
If forecasts of city population growth make use of data on urban fertility and mortality, how should we forecast these demographic inputs? Is there any basis on which internal migration can be forecast? What methods allow the spatial extents of city populations to be forecast? What has been learned from the past? Has the spatial extent of cities ever been linked to specific elements of demographic change? Should we consider investing in retrospective analyses that are generalizable rather than city-specific? What measure of uncertainty would be associated with spatial forecasts? What sorts of checks between city-specific and national-level forecasts would be desired.

The panel presentations included some new and critical views: Mark Montgomery presented on classical and Bayesian approaches to forecasting city populations. Brian O'Neill, of the International Institute for Applied Systems Analysis, presented multi-state projections of urbanization and migration in a case study of China. Philippe Bocquier of AFRISTAT in Mali took a critical overview of recent advances in urban forecasting. Lastly, Eduardo Moreno, of the Global Urban Observatory of UN-HABITAT discussed projecting the urban poor. Each presentation used a new approach demonstrating that even conventional and currently-available data inputs can be used effectively to inform urban population forecasts.

It was evident that new methods for establishing uncertainty and for determining counterfactuals will be necessary. It was further shown that spatial information can be incorporated into forecasts, in more than one way. Thus, while the techniques discussed are not yet well tested, they appear to be highly promising.

Institutional Investments in Data: Urban Projections, Poverty, and the MDGs
The final panel in the workshop discussed the need for investments in data and was informed by the following leading questions: Which data sources are relevant but are still under-utilized? What are the existing and emerging non-conventional data sources that should be exploited? What investments are required to link spatial data and tabular information from censuses? In terms of administrative area building-blocks, what size is ideal for urban areas? What can be expected from the next global census round? Will the transition to rolling census systems have a significant impact in the field?

Presentations in this session included those from Dr. Rogelio Fernandez, of UNFPA, Dr. Tukufu Zuberi, of the University of Pennsylvania, Ms. Livia Montana of the Demographic and Health Surveys of Macro International, and Dr. Andy Nelson of the Joint Research Centre of the European Commission. The types of data discussed included census, surveys, and satellite data. Each presentation made it clear that although advances are continuously being made, data collection and dissemination with an explicit urban perspective and relevance is still largely missing or incomplete. There was discussion on how current surveys be made more informative about the urban landscape. Although no consensus was reached, there was discussion on the merits, and expense, of over-sampling urban and slum dwellers for the purpose of monitoring MDG, urban poverty, and future population projection. The issue of data ownership was also discussed.

It was evident that investments into census and survey data streams should not be taken for granted. Census-taking and dissemination of census results are often postponed due to financial constraints. A lack of appropriate technical assistance is another constraint especially in resource-poor countries. DHS data are collected and disseminated through investments from the US Agency for International Development, and in some respects that agency’s priorities have a bearing on which countries are selected for these surveys as well as the country-specific content.

Satellite data are generally not collected with social science applications in mind. How to improve urban detection through future generations of satellite sensors, in a manner that would be useful to social scientists, received considerable discussion. A National Research Council study (i.e. the Decadal Study), which is currently underway, was discussed and the participants discussed how they could bring spatially-informed social science views into that process.

For all types of data, it was clear that data collection and their downstream uses need to be considered in tandem. This is particularly salient for understanding urban areas for which the current data streams are still in important ways limited.

Workshop Outcomes
Several goals of the workshop, ranging from short to long-term, were met. An immediate goal was to foster dialogue between agencies and researchers. In the wake of the workshop, there has been an increase in dialogue among several UN agencies (the UN Populations Division, Statistics Division, UN-Habitat and UNFPA) and various participants based at research institutions, as well as among individual participants.
Another short-term goal was to prepare a letter to the National Research Council’s Decadal Study Panel (committees on “Application”, “Land Use”, and “Health”). This letter, prepared by a subset of the workshop’s participants, was used to formulate their recommendations to be made public in the summer of 2006. A copy of this letter is available here.

A medium-term goal is to produce a volume of papers representing recent advances discussed at the Workshop. The book would tackle some of these key interdisciplinary issues. The organizers are still developing the structure of the volume. It is anticipated that an additional, smaller meeting will be required to bring key contributors together.

The longer-term goal was to revise the basic methods for estimating and projecting urban and city populations. In this respect, the organizers submitted a proposal to the US National Institutes of Health to develop these new methodologies.