Bernhard Babel
University of Cologne
Reliable forecasts of life expectancies are of paramount importance for the financial stability of social security systems and the life insurance industry. A discrete-time and a continuous-time stochastic process are proposed in order to model the dynamics of German mortality rates from which life expectancies are calculated. More precisely, a panel data model is utilized which distinguishes between a common time effect and a common age effect. The model is easy to fit, yields interpretable parameters, and allows for a simple analysis of the forecast error. The main applications of the model are the forecast of mortality rates - and the resulting life expectancies - and the pricing of mortality
derivatives.
Email: babel@wiso.uni-koeln.de
Using age and spatial flow structures in the indirect estimation of migration streams.
James Raymer1 and Andrei Rogers2
1University of Southampton, 2University of Colorado
This paper outlines a formal model-based approach for inferring interregional age-specific migration streams, in settings where such data are incomplete, inadequate, or unavailable. The estimation approach relies heavily on log-linear models, using them to impose some of the regularities exhibited by past age and spatial structures or by those obtained by combining and borrowing information drawn from other sources. The approach is illustrated using data from the 1990 and 2000 United States and Mexico censuses.
Email: j.raymer@soton.ac.uk
Evolution of sex differentials in mortality among Israeli Jews
Laura Staetsky
University of Southampton
This study explores patterns of sex differentials in mortality exhibited by the Jewish population in Israel in the second half of the 20th century in relation to other countries in the world. The sex differentials among Israeli Jews have been significantly lower than in countries of Europe and North America. They are currently widening over time at a very slow pace, whereas in Europe and North America, the differentials have been narrowing since around 1980.
To study these patterns, multiple data sources are used. The main sources of information come from the Israeli Central Bureau of Statistics and the Human Mortality Database provided by the University of California at Berkeley in the United States and Max Planck Institute for Demographic Research in Rostock, Germany.
These data allow the relative positioning of Israeli Jews in the world context using cross-sectional and longitudinal analyses of age-specific sex differentials in mortality. The study attempts to establish whether or to what degree sex differentials among Israeli Jews has been exceptional and also identify major sex and age -specific features of Israeli Jewish mortality responsible for Israeli Jews' positioning. The paper ends with a discussion on different socio-cultural backgrounds among Israeli Jews to identify compositional factors, and on sex differentials in mortality among Jewish communities living elsewhere in the world.
Email: lst@soton.ac.uk
Household projections in England.
Dave King
Population & Housing Research Group, Anglia Ruskin University
This paper examines the latest DCLG (formerly ODPM) 2003-based sub-national household projections for England and compares these with the previous 1996-based set.
The projections are based on various components, most of which are exogenous at the national level, at least. A key endogenous process is the projection of household representative rates, the trends for which are drawn primarily from the 1971, 1981, 1991 and 2001 Censuses of Population. These rates allow the projection of household composition as well as the total number of households. The paper considers some issues relating to the projection of these household representative rates.
Email: d.g.king@anglia.ac.uk
Esther Roughsedge
General Register Office for Scotland
This presentation describes the results of the latest household projections for Scotland, which were published in May 2006. Over the next 20 years, Scotland's population is projected to increase by around 1%, but there is a far larger projected increase in the number of households. This is mainly due to the population ageing, and more people live alone or in smaller households.
This presentation also describes how the household projections for Scotland are produced, as this differs from household projections for other parts of the UK. Over the next two years, the General Register Office for Scotland will be reviewing the household projections methodology, and this will be discussed.
Email: esther.roughsedge@gro-scotland.gsi.gov.uk
Robin Edwards
Hampshire County Council
This presentation describes the Small Area Population Forecasting model developed by Hampshire County Council. This is a short term dwelling supply led model, which produces population forecasts by single year of age and gender, and dwelling stock forecasts for each of the 5,400 2001 Census output areas comprising the County of Hampshire and the unitary authorities of Portsmouth and Southampton.
The model is linked to the County Council's Land Availability Monitoring System which provides detailed information on all planning consents for residential development in the County and allocations in development plans, and monitors starts and completions on each site, including dwelling stock losses.
The information held includes details of dwelling type and tenure, and number of bedrooms.
The population model algorithm converts this information into forecasts of population change resulting from new development, and also incorporates forecasts of natural change and the net effect of population in- and outflows within the existing stock of dwellings.
The presentation will describe the model algorithm, with particular reference to the dwelling stock change module, and briefly discuss some of the uses to which the information has been put.
Email: robin.edwards@hants.gov.uk
Anatole Romaniuc
University of Alberta
NB. Due to unforeseen circumstances, this paper will not now be presented at the Conference, but a full-text PDF can be accessed from the title line.
Insufficient attention has been paid to the question of what we mean by forecasting as applied to demography. The paper begins with the taxonomy of future-oriented endeavours, according to their perceived goals, their methodology and underlying assumptions, as well as their configurations. Distinction is made between prospective analysis versus prediction, process-oriented versus goal-oriented (normative) projections, single-purpose versus multi-purpose large-scale and multidi¬mensional projections.
This opening discussion leads to epistemological considerations involving such propositions as:
1. The future is seen as the outcome of necessity, will and chance.
2. Inherent creativeness of the future rather than its discovery. In other words, projections as tools not so much for predicting and discovering the future than for creating and managing it.
3. Analytical credibility as requirement for projection acceptability.
4. usefulness as validation criterion.
5. public recognition.
The conclusions place emphasis on prospective analysis rather than on prediction; on analytical credibility rather than on predictability; on impact rather than accuracy as forecast validation criterion; projection as tool of creating or managing rather than discovering the future.
Email: romaniuc@netrover.com
Population projections and growth management in Miami-Dade County, Florida
Oliver Kerr
Miami-Dade County
Miami-Dade County, Florida, is a rapidly growing metropolitan area with a 2005 population of about 2.4 million persons. For 25 years the County government has used a time-phased urban development boundary coupled with officially adopted small-area population projections to manage the urban development of the County.
The projections are made in two phases.
-
First, Countywide long-range projections are made using a standard components-of-change model. Birth and death rates are extrapolated to project future resident births and deaths. Gross migrations flows (in and out) are also projected for both domestic and foreign populations.
-
Second, population projections are made for 32 statistical areas of the County using logistic curves which incorporate current and historic curve combined with an estimate of the future population capacity of each area. The estimate of future capacity is based on the planning or zoning capacity of vacant land and an estimate of the redevelopment capacity of each area. The small-area projections are updated as necessary, roughly every two years, and are controlled to the County totals for future years.
The County has found these techniques to be a useful way to manage urban growth of about 30,000 persons a year.
Email: kerr39@bellsouth.net
Paul Voss and Guangqing Chi
University of Wisconsin
Existing demographic forecasting techniques for small areas often are not guided by theoretical models or demographically-driven conceptual schemes (such as exist for larger areas). In this study we test a theory-driven, spatio-temporal approach to small-area population forecasting that goes outside the world of traditional demographic inputs to "improve" the forecasts.
The basic forecasting framework - multivariate spatial regression - is not much used by applied demographers for forecasting these days, and it thus seems ripe for testing. In particular, we develop five, mostly non-demographic, indices (sustainability, liveability, accessibility, developability, and desirability) and fit a spatio-temporal regression model to examine population change at the minor civil division (MCD) level of geography in Wisconsin (U.S.A.). In a series of four increasingly complex models and for each MCD, the population growth rate for 1980-1990 is regressed on (1) its growth rate for 1970-1980, (2) several independent variables in 1980, (3) neighbourhood characteristics in 1980, and (4) neighbourhood growth rates for 1970-1980. The estimated coefficients and spatial parameters are then used for projecting population in 2000.
The accuracy of the forecasts is assessed by comparison against 2000 census counts, and the utility of the spatio-temporal approach is judged by comparing our forecasts with simple small-area extrapolation forecasts for MCDs, which we posit to be the present state-of-the-art.
Email: voss@ssc.wisc.edu
Projecting small area populations by ethnic group: Challenges and solutions.
Paul Norman1, Ludi Simpson2 and Abdelouahid Tajar2
1University of Leeds, 2University of Manchester
In rapidly changing urban systems it is necessary to project the impact of emerging populations but there are invariably few data to estimate their demographic trends. This general problem of projecting the size and structure of small populations is addressed here in the context of neighbourhood housing needs at sub-district level in northern English towns where new Asian populations are growing and suburbanising.
Alternative schemes are tested for a variety of geographical scales, ethnic classifications, age-specificity of demographic rates, methods of imposing constraints and avoidance of rates based on small populations. Alternatives add or reduce the volatility of results to varying degrees. The avoidance of rates based on small populations, however, risks bias in population projections since the emerging characteristics of new populations may be ignored.
Email: p.d.norman@leeds.ac.uk
Harvey Snowling
General Register Office for Scotland
Each year, the General Register Office for Scotland (GROS) produces mid-year population estimates at local authority level. Following a demand for population estimates at a smaller geography level, the cohort-component method used to create the mid-year local authority estimates was adopted for the production of small area population estimates (SAPE). Using the Scottish Neighbourhood Statistics 'datazone' as the level of geography for SAPE , GROS has published population estimates for quinary age groups for the years 2001 to 2004 for each of the 6,505 datazones in Scotland.
Following the publication of these estimates in October 2005, GROS has undertaken a quality assurance exercise that involves comparing the SAPE data with those from a number of administrative sources, such as the ScotXed School Census, the Child Benefit database and the Older Persons database. This paper looks at the initial results of the SAPE quality assurance investigation, with reference to any problem areas that have been identified and possible improvements to the method used to calculate SAPE.
Email: harvey.snowling@gro-scotland.gsi.gov.uk
Simon Brown
Hampshire County Council
Over the past year we have updated our population forecasting model to provide estimates by census output area, rather than by the old enumeration districts. The first step was to produce estimates for the 2001 residential household population broken down by output area, year of age and gender. We also produced estimates of the number of students and members of the armed forces and their dependants in the private household population. This paper will explain how we dealt with limitations of the census data, including disclosure control on low values and often limited data at output area level, to produce the best possible estimates as a base for our model. It will also summarise how we implemented our revised model so that it could efficiently handle the very large amount of data involved in producing population forecasts at this level up to the year 2012. The detail of the model's structure will be covered in the paper by Robin Edwards.
Email: simon.brown@hants.gov.uk
Stanley Smith
University of Florida
Prediction Intervals for County Population Forecasts. Population forecasts entail a significant amount of uncertainty, especially for long-range horizons and for places with small or rapidly changing populations. This uncertainty can be dealt with by presenting a range of projections or by developing statistical prediction intervals based on models that incorporate the stochastic nature of the forecasting process or empirical analyses of past forecast errors. In this paper, we develop and test empirical prediction intervals for county population forecasts in the United States. We find that prediction intervals based on past distributions of forecast errors provide reasonably accurate predictions of future forecast errors. We believe empirical prediction intervals can give data users valuable information to consider when they use population forecasts to plan for an uncertain future.
Email: stans@bebr.ufl.edu
Towards measuring uncertainty in population data generated by the cohort-component method.
David Swanson
University of Mississippi
This paper empirically examines a system proposed by Swanson et al. for placing a formal measure of uncertainty around population projections made using the cohort-component method. The measure is in the form of a Mean Square Error Confidence Interval, which takes into account bias and random error. The bias portion is developed using Demographic Analysis estimates of net undercount error for age-race-sex groups over two successive census counts. The random error portion is developed using variation in mortality. Together, these two sources of error can be extended to cover the uncertainty in estimates and short-term forecasts generated by the cohort-component method, given certain restrictions regarding strategic assumptions. The confidence intervals are intended to provide a set of boundaries for both the total population and age-race-sex groups once a given set of strategic assumptions is established regarding secular trends in the components of population change. The system is illustrated with a set of intervals around age and sex groups projected for a small area, Nye County, Nevada. Strengths and weaknesses of the system are discussed.
Email: dswanson@olemiss.edu
Richard Belding
Aberdeenshire Council
The objective of this presentation is to compare and evaluate historic projections and forecasts of population for two council areas (Aberdeen City and Aberdeenshire) in the NE of Scotland, and also for the Structure Plan Area, which is formed by combining these two areas.
When compared with other Scottish council areas, Aberdeen City is near the negative end and Aberdeenshire is near the positive end in terms of ranked percentage change between the General Register Office for Scotland (GROS) 2002 and 2004 based population projections for 2018. The Structure Plan Area occupies a more intermediate position, but below the Scottish level.
In order to examine the background to this situation more fully for areas in the NE of Scotland, GROS population projections, which were produced every two years from a 1992 base, are compared with each other, and with the original and revised annual mid-year estimates of population also produced by GROS. This analysis displays some of the uncertainty involved in these projections.
A similar approach is also taken with the four strategic forecast updates produced for the same areas by the councils in NE Scotland since local government re-organisation in 1996.
The recent population projections and forecasts are then compared for the same areas. The reasons for the differences are discussed.
Finally for these same areas, the 'accuracy' of projections and forecasts produced at around 1981 (the latter using two different methods) is examined by comparison with the eventual mid-year estimates of population.
Email: richard.belding@aberdeenshire.gov.uk