Household, Income and Labour Dynamics in Australia (HILDA) was commissioned by the Commonwealth department currently known as Families, Housing, Community Services and Indigenous Affairs (FaHCSIA, the Department), previously known under earlier administrative arrangements and perhaps still better recognised as the Department of Social Security. FaHCSIA owns the data and licenses their use by outside researchers. The survey itself is managed by the Melbourne University Institute for Applied Economic and Social Research (the Melbourne Institute). A team of expert researchers from the
Melbourne Institute designs the questionnaire for each annual wave in consultation with the Department, which must approve all questions, and an advisory panel representing key government and academic users. The field research for the first seven waves was carried out by the social research firm AC Nielsen Ltd.
HILDA was developed to address a gap in Australia’s infrastructure of social statistics. While the ABS has built up a thoroughly comprehensive repertory of time-series survey data based on cross-sectional samples, the only longitudinal surveys previously available with a panel sample were restricted in their coverage to fairly specialised populations or ran for only a few years. Panel surveys use the same sample for each run and consequently make it possible to track the experiences of individuals rather than just net change within populations over time. Such individual-level data are of particular interest to FaHCSIA because they make it possible to map the dynamics of the social processes leading to or resulting from the kinds of life event that trigger, or are affected by, its activity, e.g. marriage, family formation, separation, job loss, labour force entry and exit, and episodes of ill-health (Wooden and Watson 2007: 209). In this respect they assist the Department in forecasting its future workload, evaluating its programs and identifying emerging policy challenges.
The scope of the survey is very broad, reflecting the range of life events and social developments to which FaHCSIA’s programs need to respond. Aside from the
Department’s own needs, the dataset was designed from the start to meet the requirements of other social researchers working in related areas of interest, and as such overlaps the compass of more specialised existing surveys. The three main designated subject areas are
household and family dynamics; income and welfare dynamics; and labour market dynamics (Wooden and Watson 2007: 210). A large part of the dataset consists of extensive and detailed information on household and individual income and expenditure. The core set of questions which are asked in unchanged format from year to year covers nine key areas:
(i) education;
(ii) employment status;
(iii) current employment (location, type, working conditions and job satisfaction); (iv) labour market experience of persons not currently employed at the time of
interview;
(v) detailed calendar of labour market and educational activity over last 12 months; (vi) income;
(vii) family situation (including non-resident children); (viii) partnering and relationships;
(ix) living in Australia (miscellaneous topics, e.g. disability, life satisfaction, housing, caring responsibilities).
(Wooden and Watson 2007: 211)
This core questionnaire has been progressively expanded to cover new issues which are intended to be permanent features of the survey: for example, employed respondents have been asked about their work-based training experience since Wave 3, and the set of questions on job quality and characteristics has been substantially extended from Wave 5 onwards, as further discussed in section 3.2 below. In addition, special modules have been inserted in individual years to gather more detailed data on issues of current interest; so far they have included retirement intentions and income, personality traits, obesity, health insurance, household wealth, youth and fertility. Some of these are unique to a single wave, while others are scheduled to be repeated every few waves. The scope of the questionnaire is intended to evolve in line with policy interests: for example, future waves are expected to contain an increasing emphasis on health issues (Wooden and Watson 2007: 228).
HILDA forms part of a growing body of international panel data collections covering similar topics, of which the most important are the British Household Panel Survey and the German SOEP. The questionnaire has been purposely designed to provide compatible data with these surveys (Haisken-deNew and Hahn 2006).
The original panel was set up in 2001. The primary sampling unit is the household, defined as “a group of people who usually reside and eat together”, following the standard ABS definition (Watson N 2008: 110). The sampling method can best be described as a
geographic cluster sample. A sample of 488 census collection districts was drawn out of a total of some 38,000 across Australia, stratified by population, State and metropolitan/ non- metro region. Within each of these districts a sample of between 22 and 34 dwellings was selected. Where a dwelling was occupied by multiple households, a maximum of three households were sampled. The resulting initial sample included 11,693 in-scope
households, of which 6,872 provided full responses and a further 810 part responses. In this first wave 19,914 persons were enumerated and 13,969 provided interviews (Watson N 2008: 117).
In principle members of the panel, once recruited, remain continuing sample members (CSMs) indefinitely. Children who are born to or adopted by a CSM, and anyone who has a child with a CSM, are recruited with the same status. Other ways of being recruited are to move into a household that forms part of the sample, or to be a member of a household to which a CSM moves, but recruits in this category are counted only as temporary sample members (TSMs) and not tracked if they subsequently leave the household. The only ways a CSM can pass out of scope are to die or to move permanently outside Australia. In practice, the composition of the panel has been far more significantly affected by attrition, a matter that is addressed in more detail in 4.3.2 below. However, considerable trouble is taken to track members who are initially uncontactable, and even if a member eventually cannot be contacted or refuses to participate in one wave, efforts will continue to include him/her in subsequent rounds of data collection.
The survey is conducted annually, generally between August and December, though the fieldwork can be extended into the following March for hard-to-track sample members (Wooden and Watson 2007: 212). This means that the gap between surveys can extend in some cases to as much as 18 months; several of the variables have alternative items in the dataset which have a reference period limited to the 12 months preceding interview. The questionnaire is made up of three parts. A household questionnaire, administered to one member of the household, includes questions covering all members of the household (“enumerated persons”). This is followed by an individual questionnaire for each
household member over the age of 15 (“responding person”). The latter has separate versions for new and continuing sample members. These modules are administered face- to-face by an interviewer, though in practice a small but growing proportion of interviews in the later waves has needed to be carried out by telephone because of difficulties in making face-to-face contact (Wooden and Watson 2007: 212). The interviewer survey is supplemented by a self-completion questionnaire (SCQ) which is left with each responding person to be returned by mail or subsequently collected by the interviewer. The SCQ contains some questions which are difficult to administer in real time, e.g. because of the need to consult financial records, and some which are considered so sensitive that
respondents would be reluctant to answer them frankly face-to-face or in the hearing of other household members. Data items in the published dataset are classified under the questionnaire on which they originated, but in many cases are also merged to form composite derived variables for each responding person.
Six waves of data have so far been made available for analysis, running from 2001 to 2006. New waves of data are generally released in the November following the year in which the fieldwork was undertaken. Although initially funded only for four waves, the survey has so far received new appropriations each time its existing one has run out, and new funding announced in the 2007 Budget has ensured that a minimum of twelve waves will be conducted (HILDA Annual Report, 2007).
FAHCSIA permits other Commonwealth and State agencies and academic researchers to use the dataset on application, under an individual or institutional Deed of Licence. The Deed is issued only for a limited period and permits use of the data only for the purposes specified in the application, and subject to stringent controls on the dissemination of unit record data. These controls are considered necessary to protect respondent privacy, in view of the extremely sensitive nature of some of the questions asked (e.g. on individual and household finances, health matters and respondents’ perceptions of the quality and
durability of their personal relationships). All but the most secure users are given access only to a public release file which contains some confidentialised items and limits
disaggregation in order to minimise the risk of identifying individuals. One aspect of the latter which is particularly relevant to the research carried out for this thesis is that data on respondents’ occupation and industry of employment are available only down to the 2-digit level.