Stories from the South – Longitudinal studies from Australia and New Zealand 1 |
|
| Session Organisers | Ms Joanne Corey (Australian Bureau of Statistics) Professor Susan Morton (Centre for Longitudinal Research - University of Auckland) |
| Time | Friday 19th July, 11:00 - 12:30 |
| Room | D25 |
Australia and New Zealand have a wealth of wonderful longitudinal studies. The range of topics and target populations is broad and includes:
• Babies and Children and their Parents
• Families
• Young People
• Women’s Health
• Men’s Health
• Ageing
• Indigenous Children
• Migrants
• And many many more
This session is interested in hearing from survey methodologists and practitioners from Australia and New Zealand who work in this area and would be interested in sharing their stories.
We are also keen to hear from researchers from around the world who use Australian and/or New Zealand longitudinal datasets to hear about the work they are doing.
For example, we would love to hear about:
• data collection activities and methodologies
• engagement strategies and incentives
• communication with respondents including initial approach, between waves, etc.
• data linkage including consent processes
• strategies employed to reduce panel attrition
• engaging and novel methods of relaying study results back to participants
• social media strategies
• multi-mode experiences
• re-engagement of past non-responders
• using Australian/New Zealand longitudinal data together with other longitudinal data sets
Keywords: longitudinal, engagement, survey methods
Ms Nicole Watson (University of Melbourne) - Presenting Author
Attrition of sample members from a longitudinal survey can undermine the quality of the data and its research potential, especially when the sample members who drop out are different from those who do not. People who move house are more likely to drop out of the survey as they are harder to locate, and once located, may be harder to interview in the remaining fieldwork time available. Moving coincides with many other life events (such as changes in marital status, the birth of a child, buying a home, changes in employment, or retirement) and if movers are not adequately interviewed, this may result in the study under-representing these changes and the events that occur after a move. This paper examines the weighted estimates of the rate of moving by age in a long running household panel study, the Household, Income and Labour Dynamics in Australia (HILDA) Survey, compared to official cross-sectional data sources and probabilistically linked Census data. Geographic mobility is examined over one-, five- and 10-year periods. Some of the differences that occur in the mobility estimates is a result of item non-response or recall error in the cross-sectional sources but little evidence is found of the differential impact of attrition in the HILDA Survey. There is, however, some indication that the longitudinal survey data underrepresents long distance moves. Other differences between the data sources are investigated by fitting logistic regression models of mobility to estimate the effect of housing tenure and education levels over the life course. These models show similar overall trends, but there is some evidence of differential effects for renters with lower education levels which may be due, at least in part, to the differential role recall error plays in these measures.
Dr Galina Daraganova (Australian Institute of Family Studies) - Presenting Author
Mrs Dinusha Bandara (Australian Institute of Family Studies)
Dr Mabel Andalon (Australian Institute of Family Studies)
The Australian Longitudinal Study on Male Health, also known as the Ten to Men, is the first national longitudinal survey that focuses exclusively on male health and wellbeing. The study covers physical and mental health, health-related behaviours, health literacy, health service use and the social determinants of health. The study aims to build the evidence base to inform the development of health policy and programs targeted to the changing needs of boys and men.
The sample design for Ten to Men was a stratified, multi-stage, clustered random sample design with separate cluster samples taken from each regional stratum. The study commenced in 2013 with around 16,000 males participating in the first wave of data collection (7% boys aged 10-14 years, 6% young people aged 15-17, and 87% men aged 18-55 years) and around 12,000 males participating in second wave of data collection that took place in 2015/2016. In the second wave response rate was 83% for boys, 73% for young men, and 76% for adults.
This paper will provide an overview of the study’s cohort profile focusing on sampling and recruitment methodology. It will discuss challenges identified at the end the recruitment such as lower than anticipate response rate at wave 1 and sample representativeness by age groups and localities. Lastly, it will discuss different options of cohort top-up and reconciliation.
Professor Deborah Loxton (University of Newcastle) - Presenting Author
Ms Natalie Townsend (University of Newcastle)
In 1996, 42,000 Australian women in three cohorts born 1921-26, 1946-51, and 1973-78 were recruited to take part in a longitudinal study on women’s health. In 2012-3, another 17,000 women born 1989-95 joined their ranks to comprise the largest and longest running study on women’s health in Australia. Despite their shared common purpose to inform women’s health and health policy, the four cohorts who make up the Australian Longitudinal Study on Women’s Health (ALSWH) were recruited differently and are subject to tailored, dynamic retention methods.
The advent of mass market research, online surveys, commercialised survey data collection and online incentives for participants has permanently changed the research environment. Where once we could expect recruitment rates of between 50-80%, now some population groups are lucky to yield 6%. To address this drop in recruitment rates, an innovative approach was necessary. Therefore, the 1989-95 ALSWH cohort was recruited using open recruitment, including strategies such as Facebook, other social media, referrals, traditional media and a fashion promotion.
Of the 17,000 women who completed the baseline survey, around two-thirds completed the second survey, and just over half completed the third and fourth surveys. Women did not complete surveys consistently, only 38% completed all surveys. Retention was associated with age, education, health, and health behaviour. However, recruitment method was also a determinant of study participation. While women were more likely to be recruited via social media than traditional methods, retention was higher for women recruited through traditional media and referrals. A balance between efficient, affordable and effective cohort recruitment must be struck with the potential drawbacks of attrition related to particular recruitment methods. In the third millennium, we need to develop our understanding of the motivations that lead people to participate and to continue participating (or not) in cohort studies.
Dr Steven McEachern (Australian Data Archive) - Presenting Author
Mr Sebastian Kocar (Australian Data Archive)
Anonymization of survey data collected from the same respondents at different points in time, such as online panel data, is a special case in the area of statistical disclosure control. While online panels play an important role today in survey research, data archives disseminating such data have to face particular privacy restrictions that result from the capacity to follow individual participants over time. Traditional SDC methods assume the released data to be static, but they are not in the case of online panels with several waves of data collection. Disseminating panel data could increase disclosure risk as the abundance of demographic information included in the original data enables the potential linking data of different waves. This accumulation of identifying information, and increasing potential disclosure risk, has the subsequent potential to reduce of trust in the data producer/depositor over time. To address this problem, we developed a procedure using traditional SDC methods for decreasing disclosure risk while selecting a limited number of demographic variables in the released data. We used the following selection criteria: importance for analysis, contribution to global identification risk, availability of background information, and the effect on data utility. We will present the results of the anonymization of several waves of Life in Australia panel data from both SDC perspectives - disclosure risk and data utility.