ACSPRI Conferences, ACSPRI Social Science Methodology Conference 2010

Font Size:  Small  Medium  Large

Patterns and correlates of unit non-response for Waves 2 and 3 of Growing up in Australia: the Longitudinal Study of Australian Children (LSAC)

Gamini Wijesekere

Building: Holme Building
Room: Common Room (Boardroom 418)
Date: 2010-12-01 01:30 PM – 03:00 PM
Last modified: 2010-11-17


While non-response is common to all social surveys, non-response is particularly a problem in longitudinal studies where the same respondent is interviewed a number of occasions over a period of time. If some eligible respondents do not respond in subsequent waves of a longitudinal survey and the characteristics of persons failing to respond are different to those who respond to the survey such failures could lead to bias in some survey variables. Although high response rates do not necessarily guarantee high data quality they nevertheless are a useful gauge. In longitudinal surveys nonresponse accumulates over time and may cause the survey sample to gradually become unrepresentative of the target population. Consequently it has become important to evaluate the pattern of nonresponse in each wave of longitudinal surveys and identify any socio-demographic and economic characteristics of those who are likely to drop out of the surveys.  Nonresponse in longitudinal surveys arises when eligible respondents refuse to cooperate with the survey (refusal) or the respondent is not available when the interviewer attempts to contact them for an interview (noncontact).

The main purpose of this research is to examine in detail the unit nonresponse to Waves 2 and 3 of the Growing up in Australia the Longitudinal Study of Australian Children (LSAC) by primary parents and to highlight any socio-demographic and economic characteristics of those who did not respond to the two survey waves. Existing studies of nonresponse in LSAC focused on overall nonresponse rates without differentiating the unit nonresponse by its source—noncontact or refusal by the eligible respondent. This distinction is important as the strategies to minimize the nonresponse could well be different according to the source of nonresponse. The knowledge of these helps to develop fieldwork strategies to minimize dropouts of primary parents from future surveys and to assist in designing weighting strategies to ensure the representativeness of the future waves of the survey.

LSAC is the first national study of young children conducted in Australia and if funded by the Department of Families, Housing, Community Services and Indigenous Affairs (FaHCSIA). The survey is currently conducted jointly with the Australian Institute of Family Studies (AIFS) and the Australian Bureau of Statistics (ABS). Its aim is to provide data to contribute to a deeper understanding of child growth and development in the current social, economic and cultural environment.

The main source of data for the study is the information provided by the primary carer of the study child to the survey interviewer at a face-to-face interview. The study used two cohorts of children: infant or B cohort (infants born between March 2003 and February 2004), and child or K cohort (children (4-5 years olds, born between March 1999 and February 2000). The first wave of data collection was completed in 2004 and covered 10,090 (5,107 infants and 4,983 children 4-5 years old) primary parents. In Wave 2 conducted in 2006 (interviewed a total of 9,069 primary parents: 4,067 infants and 4,462 children). In Wave 3, conducted in 2008, the number of primary parents interviewed dropped further to 8,718 (4,387 and 4,331 respectively for infants and child cohorts).

The present research will examine by multivariate methods the nature, extent, and profile of nonresponse in Waves 2 and 3 according to the source (type) of non-response: noncontact and refusal. Socio-economic characteristics examined in the research are those reported at Wave 1. For the research, the unit nonresponse is defined as the eligible primary respondents who participated in Wave 1 but failed to respond to a subsequent wave, either due to the interviewer failing to contact the respondent or the respondent being contacted but refusing to participate.

As the dependent variable consists of three categories—response, noncontact and refusal—multinomial logit models have been applied to the data. The models fitted will examine socio-economic characteristics of primary parents according to the following combinations of response/nonresponse: response vs refusal and response vs noncontact. Multinomial logit models are fitted separately for the nonresponse in the infant cohort and the child cohort. Explanatory variables for the research are chosen based on a literature review and using the framework for survey response/nonresponse proposed by Cooper and Groves (1998) as a guide.

The models will examine the socio-economic, household and neighbourhood characteristics of primary parents explaining the probability of their being contacted by the interviewer and then, once contacted, the decision by the  respondent to participate or not participate in the survey. Any variations in the characteristics between the B-cohort and K-cohort parents will be highlighted. The research also discusses the limitations of the data to understand the propensity to respond/non-respond and proposes that that information on interviewer characteristics (i.e. age, sex, previous survey experience) and details of the field work strategies (interviewer calling strategies, the number of call-backs as well as reasons for noncontacts and for any opt-outs) need to be collected and available in the public domain which will enhance our understanding of the pattern of noncontact and refusals.