Demography & policy strand abstracts


UK Censuses and methodology: Monday 12 September 16:45pm 

Scotland's Census 2021: Developing content following topic consultation 
Kirsty Maclean, Cecilia Macintyre, National Records of Scotland

 National Records of Scotland carried out a consultation on the topics to be included on the 2021 questionnaire which ended in January 2016. NRS will publish a response detailing the results and outlining a range of activity and further engagement to shape the plans for developing a set of questions to be submitted to the Scottish Parliament. The detail of the content of the questionnaires will only be finalised after a comprehensive programme of research, testing and further discussion with users. This talk will highlight the results of the consultation and will illustrate the range of issues which will be addressed through the research programme, using the topic of language as an example. 

What is your occupation? – Working towards online self-identification of occupation for the 2021 Census for England and Wales 
Scott Redgwell, Office for National Statistics, Paul Waruszynski, Office for National Statistics 

The capturing, coding and processing of variables, with a large number of potential responses and a large coding index, have historically been time-consuming, complex and expensive to administer for an operation such as a national census. This presentation will focus on occupation; we will detail how this variable was captured and coded for the 2011 Census, and explain the challenges and problems and other drivers for change, including the 2021 Census being ‘online first’. This presentation will then explore other developments within this field, looking at what else has been happening within ONS, academia, and best practice internationally. This will lead to the exploration of different options being considered by the office for the collection of occupation for the 2021 Census. These options include: 1. Improving the quality of automatic coding processes, for example, by asking supplementary questions where the initial response is too vague. 2. Employing a ‘coding tree’, allowing respondents to work through a full occupation hierarchy to self-identify 3. Adopting a semantic coding approach, dynamically displaying options to the respondent from the coding index based on what they have typed. Each option will be assessed based on the impact they have on the quality of the data collected, the impact of the burden on the respondent and the effect they have on reducing costs. The paper will also detail how this will be tested leading up to 2021.

  Creation of synthetic microdata in 2021 Census Transformation Programme (proof of concept) 
Robert Rendell, Office for National Statistics 

The Census Transformation Programme of the Office for National Statistics in England and Wales is investigating the use of synthetic data for a number of potential uses. One potential application is the creation of a household microdata sample. Due to disclosure concerns, we have not been able to provide a 2011 Census household microdata sample accessible outside secure research environments, with sufficient utility for users, using standard disclosure control techniques. The method being tested to create a household microdata file punches holes in a microdata sample, and uses the edit and imputation process from 2011 Census (using CANCEIS) to fill in the holes. This attempts to preserve the relationships between variables within households and individuals, but introduces sufficient uncertainty to mitigate disclosure risk. Methods are being investigated to test utility and risk in the resultant data. This poster will demonstrate issues we have to overcome, and the methods we are investigating to come up with a solution to provide useful, non-disclosive, microdata for a wider range of users. Given these issues, this is just a proof of concept at this stage, to test whether the approach is feasible. 

Respondents who don’t ‘submit’ – An analysis of forced submissions from 2011 Census 
Paul Waruszynski, Office for National Statistics 

The increasing emphasis on taking censuses and surveys online allows for the collection of new types of data, which can present new opportunities to understand respondents and improve response rates. This presentation will focus on the analysis of one such set of data from the 2011 Census online questionnaire. This database contains people who started to complete their 2011 Census questionnaire online but who never pressed the ‘submit’ button. However, as respondents passed through each page of an online questionnaire, the data was still captured and used in the production of Census outputs. The analysis presented will first aim to understand this group, looking at the types of people who had their responses forced submitted. Techniques such as Principle Component Analysis and Canonical Variate Analysis have been used to identify key variables leading to sub-groups of the populations who are mostly likely to have a forced submitted response. Other analysis aims to identify whether there were specific questions where people dropped out of the questionnaire. The outcome of this analysis will contribute to identifying population groups most in need of additional help online in 2021. It will also contribute to questionnaire development by identifying the questions where they experienced difficulty. 

Small area estimation. Tuesday 13 September 09:00am 

Housing delivered in the first decade since the millennium – who lives there? A Methodology for Identifying New Residential Developments, and the Number and Ages of Residents, on Census Day 2011 at Specific Development Level for Any District in England 
Steve Clyne Educational Facilities Management Partnership Ltd (EFM) 

As part of its work forecasting the impact of the population of new developments on the local social infrastructure, and in some instances the need for additional social infrastructure, EFM has developed software to extract the relevant population data from the 2011 Census at Output Area level (that also identifies Postcodes and often Street Names). That is, the dwelling number and population data associated with homes first occupied between March 2001 and March 2011 (i.e. between the censuses). EFM has taken the software as far as it needs for its own business purposes and has no commercial interest in exploiting it. However, there may be research opportunities for others with particular expertise or research needs who may wish to explore, for example, the relationship between new dwellings in areas that are statistical neighbours, or are interested in the other Census captured data (health, ethnicity, tenure, employment and education) that can be extracted once the relevant Output Areas are identified. Illustrated talk explaining the operation of the software and a discussion of the potential for a better understanding of the impact of new housing. 

Child Benefit Data as a proxy for Resident Population Counts 
Piers Elias, Demographic Support 

Have changes to Child benefit rules led to this data source losing its value as a quality assurance tool for counts of resident children? With 2021 almost certainly being the last traditional census, ONS are seeking and testing alternative data sources while Local Authorities try to improve their estimates and projections for school children using whatever suitable data sources are available. This presentation will compare the latest Mid-Year Estimates from Mid 2015 with figures published by HMRC on numbers of claimants as at 31st August 2015. With figures available for Local Authorities by single year of age and with aggregated age groups down to LSOA level, this source used to act as a good guide to resident population numbers. Since January 2013, high income families are no longer required to register their child with HMRC and so this data source is no longer a comprehensive dataset for the under 2s and as time moves on, so the numbers missing will increase. However, there is data collected on high income families who do register and this data reveals some interesting, though not unexpected, geographical variations. 

Error in local population estimates: do the characteristics of places matter? 
Alan Marshall 1, Ludi Simpson 2, 1 University of St Andrews, 2 University of Manchester 

Local population estimates are crucial for decisions relating to local service provision. As such, the small area population estimates (SAPE) that are produced by the national statistical agencies in the UK are regularly used by a variety of public and private organisations. Yet, inevitably, we must expect error around such estimates. In this paper we examine whether this error varies systematically according to the characteristics of local places and different estimation methods. The difference between lower super output area (LSOA) population estimates (with age/sex detail) and census population counts forms our measure of ‘error’. We examine several different methods for generating local population estimates including the SAPE produced by ONS, constant population as observed at 2001, constant share of district population as observed at 2001 and a simple cohort progression model. A multilevel model is used to analyse the percentage error where the various estimates are nested within areas. We will include covariates to examine how the error varies according to area characteristics such as population size and change, deprivation, age structure and level of migration as well as the method of estimation. Initial findings from the Office for National Statistics suggest that area characteristics have important influence on the spatial distribution of error. For example, the population size of a local area exhibits an inverse relationship with % error in population estimates. Our analysis provides further evidence on the extent to which population estimates in varying types of places might be trusted outside census years. 

Population level research using the Hampshire Health Record Analytical Database
Matt Johnson, NIHR CLAHRC Wessex, Faculty of Health Sciences, University of Southampton 

When analysing health or social care data for research or commissioning purposes, we often consider our topic of interest from the perspective of one sector alone. There are many reasons for this, not least the often limited availability of data from multiple sectors. However, it is also true that no single element of the health and social care system exists in isolation, so to progress from simply describing its individual parts to understanding the whole system, it is essential that we adopt a standard working practice which considers all elements together. This is possible in Hampshire by way of the Hampshire Health Record Analytical Database; a patient-centric anonymised linked database bridging different sectors of the local health and social care system. Principally comprised of clinical data from the Hampshire Health Record shared electronic care record and augmented with data from other external sources, the database links primary care, secondary care, hospital laboratory, community care and demographic data, with an extension to include social care data anticipated in the near future. With coverage of approximately 75% of the Hampshire population (or ~1.4 million patients), the database allows for longitudinal pathway analysis at patient and population level, generating opportunities for use in clinical research, healthcare commissioning, health and social care based operational research/service evaluation, and research into the social and demographic determinants of health and inequalities. We will discuss organisational, governance and methodological issues around creating and using databases of this kind, and present some worked examples of their uses and benefits. 

Data visualisation & presentation. Tuesday 13 September 16:45pm

40 years of population studies - plus ça change? 
Jenny Boag, Falkirk Council 

My intention would be to give a slightly self-indulgent look back over the last 40 years of population work, in local government in particular. Highlights include the increasing power of computing, better availability of data, changes to the Census - some good, some less so, the development of local population censuses, the Estimating with Confidence project, neighbourhood statistics, changes to the way local government is organised from the heady days of county led research teams to the much more limited resources of today, perhaps touching on the role of consultants and where what is now called ONS fits in to all this. It would be very much a personal view which might not match the experience of other people who may have been around almost as long as I have, but I think it is worth reflecting on where we have come from to see what we have lost as well as what we have gained. 

Data Visualisation @ONS - Connecting a wider audience with Official Statistics 
Vicky Field, Frank Donnarumma, Office of National Statistics 

Early in 2015 ONS launched Visual.ons, which was aimed at delivering easily consumable content to a wide audience. Delivering statistical content in an accessible way is key to this work. The site’s motto is “official data, new light” – this captures the commitment to show fresh insights into the ONS statistics produced by our economists and researchers. This new content has reached a mass audience including those who would not have engaged with our content otherwise. This presentation will explore the reasoning behind this uncommon approach and briefly look at the skills needed to create quality content to engage with an audience who are increasingly demanding and time-pressured. It will also provide some tips for those thinking about creating their own data visualisation. The presentation will showcase a number of different examples of content and explore how different techniques, such as gamification, personalisation and topicality have been used to connect the audience with official statistics. 

Scotland’s Census 2011: How National Records of Scotland are making the most of the data 
Cecilia Macintyre, National Records of Scotland 

The National Records of Scotland used a number of approaches to increase the impact which the Census 2011 results have had. This talk will describe a number of innovative approaches which have been taken in response to user needs to supplement the traditional approach to dissemination through a website with accompanying statistical bulletins. These include - promotion of uses by others through a Census conference - short term secondments of Scottish Government analysts to develop policy focused topic reports - collaborative work with teachers' professional bodies to encourage use of Census in the classroom - working with expert users to gain insight into opportunities and barriers to using Census data. This talk will discuss how lessons learned from these approaches will be used to inform the plans for Scotland’s census 2021. 

Expanding the Greater London Authority projection model beyond London 
William Tonkiss, Ben Corr, Greater London Authority 

The Greater London Authority produces a range of population projection products for geographies within the Greater London boundary. The central model is a cohort component trend-based model which projects local authority-level populations for the 33 London boroughs by single year of age and sex. Over the last year changes in the availability of data, coupled with a need for the organisation to better understand the relationship between population change in London and the surrounding regions, has provided an opportunity to redevelop and expand this model. The expanded model generates population projections for all local authorities in England as well as national-level projections for Northern Ireland, Scotland & Wales. In addition, the newly developed model provides greater flexibility for scenario testing and the production of projection variants. In particular, the GLA has been interested in the implications of using longer term migration trends than are typically considered in the ONS sub-national projections. The model also allows for testing fertility and mortality assumptions. This presentation provides an overview of the updated GLA projection methodology, including data sources and limitations. It also includes a comparison of model results to official ONS sub-national projections and examples of the types of scenario testing undertaken by the GLA. The presentation will include a demonstration of the suite of quality assurance data visualisation tools used internally by the GLA to interrogate and understand projection outputs. 

Projections & forecasts. Wednesday 14 September 09:00am

Population and Household Projections for Scottish Sub-Council Areas (2012-based)Angela Adams 1, William Howes 1, Gail Sinclair 1, Hannah Rutherford 1, Esther Roughsedge 1, Esta Clark 1, Ludi Simpson 2, 1 National Records of Scotland, 2 University of Manchester 

This project to produce Population and Household Projections for Scottish Sub-Council Areas (SCAP) was undertaken by the National Records of Scotland (NRS) with advice from Ludi Simpson from the University of Manchester. The aim was to improve and develop the methods for producing small area projections and investigate the feasibility of producing them for small areas within all councils in Scotland. There was a demand from councils to produce sub-council projections as differences between areas within councils are not reflected in the Sub-National Population and Household Projections. Being able to identify these differences may help councils in planning for future services in these areas. The POPGROUP software package, which uses a cohort-component method, was used to calculate the population projections and a SAS-based system was used to produce the household projections. We worked closely with councils and consulted them throughout including their preferred geography. This resulted in projections being produced for 301 Scottish sub-council areas. Differences in the population and number of households between sub-council areas that are not visible in the council-level projections were observed. Several insights and opportunities for further work and consideration were discovered during the course of the project. These included improving methods for areas that contain large armed forces, prisoner or student populations as these can affect the age-sex structure of the projection. 

Pupil Forecasting for the Primary and Secondary Phases 
Heather Zawada, Hampshire County Council 

The Local Authority is under statutory duty to ensure that there are sufficient school places within their area and therefore, forecasting of pupil numbers plays a large role on decision making around building/extending and even moving provision. Predicting school place demand is a complex task. Where children go to school involves a range of different factors such as housing growth, inward and outward migration as well as parental preference. As a result, planning for school places is based on probabilities and not certainties, and whilst pupil forecasts are derived from methodology, they come without a guarantee. Alongside this forecasting, there is also a duty to respond to local need, to raise standards and promote diversity in response to government policy. Internal and external findings on the quality of schools create another angle of investigation as to whether resources are being used effectively. Hampshire County Council collects data on the past and present take up of places in all schools in Hampshire that are maintained by the local authority. This information is used together with other sources of station, principally birth and housing data, to predict the future need for school places across the county. This is what is referred to as pupil Projections" or "Forecasts". 

A pupil flow based model for school roll projections in London 
Marta Lapsley, William Tonkiss, Ben Corr, Monica Li, Greater London Authority 

The Greater London Authority produces school roll projections as a service available to London local authorities to support their schools provision planning. The growth in London’s population over the last decade has created great challenges for local authorities providing places to meet growing demand. Over the next 10 years it is expected that demand for state funded school places will increase by at least 170 thousand across London. Previous models have linked changes in pupil numbers to population growth in the borough’s school planning areas. However, in reality there is a significant amount of cross-border flow, with many schools drawing pupils from across the wider London area and beyond. This is a particularly large effect in London due to the small geographic size of local authorities. With the recent acquisition of an extract from the Department for Education national pupil database, pupil home location data can now inform the model. Pupil number projections track the GLA population projections for their home address at ward level within London, and local authority level outside of London. The model outputs projected rolls for each school by age and sex, that can be subsequently aggregated to school planning area level for the purpose of planning provision. This presentation looks at the results of explorations of this and other methods possible with datasets currently available to the GLA. 

Practical uses for a local forecasting model 
Elisa Bullen, Manchester City Council 

The Manchester City Council Forecasting Model (MCCFM) is now in its second year and has continued to evolve. The model uses small area data and local intelligence to sense-check official estimates and projections, redistributing population based on evidence and producing outputs at city and ward level for 2002 to 2026. Small area statistics have been used wherever possible to refine the resident profile. This presentation briefly revisits the origins of the in-house model, then brings things up to date, explaining how housing and electorate are now informing the output. The presentation will close with current uses of the model, focussing on how the forecasts are being used to project electors for the Local Government Boundary Review. 

New Sub-National Population Projections Method for Scottish areas 
Luke Main, Claire Crowley, Esta Clark, National Records of Scotland 

In 2015, the National Records of Scotland (NRS) commissioned academics (Phil Rees, Nik Lomax, Paul Norman, and Pia Wohland) from the University of Leeds and the York Hull Medical School (YHMS) to review the Sub-National Population Projections (SNPP) methodology used to produce Population Projections at Council area, NHS Board area, Strategic Development Plan (SDP) area, and National Park area levels in Scotland. This report (Paper 1:, along with plans for implementing many of the recommendations was jointly presented at the BSPS Conference 2015 by Luke Main (NRS) and Pia Wohland (YHMS). The primary recommendation, along with several others, was to move from a single-region net migration flow assumption to a multi-region migration rates based assumption for intra-UK moves. Since then, NRS have designed and implemented the new methodology based on these recommendations. This presentation will cover four main points: a summary of the changes made to the methodology; a comparison of the differences between 2012-based SNPP using the previous methodology and the new methodology; results for the 2014-based SNPP and reflections on the project, and future developments. 

Evidence & policy – national & international evidence. Wednesday 14 September 11:30am

  Breaking down silos around demographic data and models - experience of a city's local authorities 
Sunny Townsend, Modeller & Data Scientist, Mastodon C 

Data and modelling silos prevail in authorities and organisations that have planning needs, creating barriers to generating the information and scenarios needed for decision making. Witan is a new tool to approach city planning challenges that is being developed by data and modelling software specialists Mastodon C in partnership with the Greater London Authority (GLA), and funded by Innovate UK. Currently, London's boroughs are using Witan to carry out their own population projections based on future housing expectations, through a web-based interface and using GLA demographic models. In this talk, we present results from the use of Witan by London boroughs over the past six months and how they have used the tool in their decision making / policy making processes. We will then put this into a broader geographical context and compare how London's demographic data and models differ with other cities, such as New York. Finally, we will discuss implications for developing an international demographic modelling platform. 

Modelling the age-sex differentials influencing the severity of road traffic injuries in Oman 
Amira ALAamri, University of Southampton 

Globally, more than 1.2 million people die each year and more than 50 million are injured due to Road Traffic Accidents (RTA). The burden of RTA mortality and injuries is the highest in the Middle-East, particularly among young adults in UAE, Saudi Arabia and Oman. However, there is little systematic research on RTAs in Oman especially the extent of severity of fatal and non-fatal outcomes. This paper addresses a policy relevant question: what are the underlying effects of age and sex on the extent of severity of RTA outcomes in Oman? The severity is measured in terms of RTA outcomes ranging from fatal and serious injuries in the highest extreme, moderate and then mild and no injuries in the lowest extreme. Data for this study are drawn from the RTA database managed by the Royal Omani Police and made available for research use by The Research Council of the Sultanate of Oman. The study is based on 35,785 registered incidents. Of these, 10% were fatal, 6% serious, 28% moderate and, 37% minor injuries. About 19% were no human injuries but only vehicle damages. We applied an ordered logit regression to estimate the effect of driver's age and sex on the extent of severity of injuries, controlling for a range of incident related variables including whether the incident happened during weekday or weekend and driver characteristics. The preliminary results demonstrate evidence of significantly higher odds of severity for female drivers and those below 30 years. Keywords: Road Traffic Accident, Severity, Oman, ordered logit model. 

Population and development: what do we really know? A comparison of the recent course of human development in Mali and Ghana. 
Allan G. Hill, Social, Human and Mathematical Sciences, University of Southampton 

For years, we have assumed that there are strong causal links in both directions between demographic change and economic growth. Whether Malthusian, neo-Malthusian, Marxist or capitalist, writers have taken for granted that the two processes are connected. The links today seem less clear with the advent of the sustainable development agenda and the new Sustainable Development Goals (SDG’s). Sustainable development is different from raw per capita income growth and the indicators chosen to measure progress towards the new targets are both diverse and thin on population processes. Human development remains at the centre of most development agencies agendas but the definitions of this development include concepts as diverse as national measures of well-being and Sen’s ideas on human capabilities. To investigate some of the relationships between public policy, population processes and human development, we compare the development of Ghana and Mali since independence. Both were largely agricultural at independence in the 1960s. Then, we observe relatively small differences in per capita income and levels of living between the two countries. Today, the differences in human development between the two countries are stark. Certainly, there are differences in natural resource endowment between the two countries with contrasting colonial experiences. We argue here that the differences in living standards today, Mali still one of the poorest countries in the world and Ghana newly admitted to the ranks of the lower middle income countries, are related to public policy on demographic processes and on the quality of their respective populations, human development.