Methodological Issues in Surveying Older People 3 |
|
| Session Organisers | Mr Jan-Lucas Schanze (GESIS - Leibniz Institute for the Social Sciences (Dept. Survey Design and Methodology)) Dr Emanuela Sala (Department of Sociology and Social Research, University of Milano Bicocca) Dr Annette Scherpenzeel (SHARE – Survey of Health, Ageing and Retirement in Europe, Munich Center for the Economics of Aging (MEA) ) Professor Wander van der Vaart (University of Humanistic Studies, Utrecht) Ms Daniele Zaccaria (Department of Sociology and Social Research, University of Milano Bicocca) |
| Time | Friday 19th July, 13:30 - 14:30 |
| Room | D18 |
The fast ageing processes which European societies are going through lead to a situation that has never existed before. Many of the current pension systems in European countries will in the long term be unsustainable, the costs of health care are steadily rising and intergenerational cohesion seems to be threatened. The key to meet these manifold challenges is knowledge. National and international surveys that target on or cover the ageing population groups can deliver this knowledge but face special methodological issues. For example:
• strong systematic nonresponse related to health problems
• cognitive impairments affecting data quality
• restricted use of modes (online, mobile)
• the need of specific interviewer skills and training
• conversion of gate keepers who refuse / hesitate to let an elderly person be interviewed
• use of proxy interviews or 'triad interviews' and consequences for data quality.
• undercoverage of the institutionalized elderly
• tracing which panel members died
With this outline, we invite papers addressing any of the topics mentioned above or other issues of surveying the elderly, as well as papers presenting potential survey tools that might help to improve data collection among the elderly and reduce sources or error.
Keywords: ageing, elderly, coverage, bias, health, cognition, data quality, gate keepers, proxy, traid interviews, institutionalized population
Professor Emanuela Sala (Milano Bicocca) - Presenting Author
Dr Daniele Zaccaria (Fondazione Golgi Cenci)
In an ageing society, the availability of good quality survey data is key. In particular, longitudinal surveys of older people are very powerful research resources to study social inequalities and monitor older people’s health conditions. It is not surprising that a number of longitudinal surveys of older people has been conducted in recent decades, both in the U.S. and in Europe. The relevance of these surveys is undisputed. However, there are very few studies that systematically assessed data quality in longitudinal surveys of older people (Kalwij, 2010; Gaertner et al., 2015).
This paper aims to evaluate the quality of the survey data in the Survey for Health and Retirement in Europe (SHARE). We will assess the effect of ageing, and in particular, of changes in respondent cognitive functions, on a number of indicators of measurement error, including item non response and answers to open ended questions. Controlling for age and educational level, we expect to find a positive relationship between a deterioration in cognitive functioning and the occurrence of measurement error.
We use Wave 1, 2, 4 and 5 of SHARE data. We intend to employ a set of multilevel models and, in particular, to use growth curve models (GCM) which are appropriate statistical techniques to model change in a dynamic framework. We consider as independent variables a set of indicators of cognitive function, i.e., ability in verbal fluency and numeracy skills.
Dr Rebekah Luff (University of Southampton) - Presenting Author
Dr Amos Channon (University of Southampton)
Problems with non-random panel survey attrition are well known in relation to physical and cognitive health problems. These problems are magnified in ageing studies given many of these health problems occur more commonly at older ages. However, just focusing on chronological age (time since birth) misses the strongest association with many of these health-related variables, which is ‘time until death’. An increase in health and social care use is common in the final years of life and it is these same changes that are associated with increased attrition. This paper examines both the amount and the pattern of missing data in the final years of life and considers the implications for health and social care estimates in particular.
Using linked mortality data within the English Longitudinal Study of Ageing (ELSA) a person’s death is recorded even if they have missed several waves of data collection. This enables us to determine not just if data is missing but how much data is missing before end-of-life.
Three research questions are addressed: 1) For those who have died between waves 1 and 5 of ELSA, how many waves were missing prior to death? 2) What is the average amount of missing data at end-of-life? 3) Is the amount of missing data randomly distributed? Specifically does cause of death relate to how much data is missing?
Analysis indicates that almost 30% of those who have died have 1, 2 or 3 waves missing prior to death and the amount of missing data is not random. Using multistate life tables, it is demonstrated that main cause of death relates to the chances of having more waves missing. Those whose cause of death relates to declining function over a long period of time are more likely to have more waves missing.
Dr Marika de Bruijne (CentERdata) - Presenting Author
Dr Adriaan Kalwij (Utrecht University)
Researchers of face-to-face surveys increasingly often consider using other data collection sources to decrease costs or to enrich the existing data. In our study, we investigate how switching to alternative data collection methods affects the sample size and composition. We present the willingness of face-to-face respondents to participate in web surveys and to consent to data linkage.
SHARE, a large longitudinal Survey on Health, Ageing and Retirement in Europe, targets the general population aged 50 years and older in Europe. Our research is based on data from the Netherlands, which is one of the countries participating in the SHARE project. The survey was fielded in 2013 as personal (CAPI) interview. In addition, respondents were asked for their consent to linkage with the register data of Statistics Netherlands. In 2015, the same respondents were asked to participate in the SHARE survey offered as a web questionnaire in the Netherlands.
In this paper, we investigate the bias of the response when using web surveys or the linked micro data of Statistics Netherlands as data sources, instead of the face-to-face survey. We show which demographic subgroups are willing to switch to a web survey, or to consent to data linkage, and which groups are likely to become underrepresented in the process. In addition to demographic characteristics, we include variables related to physical and mental health, general trust, computer skills, and social and physical activity as possible predictors of the willingness rate.
The paper offers insights for researchers into the effects on the sample of the 50+ population when moving from face-to-face surveys to web surveys or using alternative data sources such as register data.