RDP 2003-08: A Tale of Two Surveys: Household Debt and Financial Constraints in Australia 3. The Data

The primary sources of data used in this paper are the Household Expenditure Survey (HES) for 1993/94 and 1998/99 and the Household, Income and Labour Dynamics in Australia (HILDA) Survey for 2001. These surveys contain detailed information on expenditure, income and demographic characteristics for households resident in private dwellings throughout Australia.[5] Both surveys collect information from all persons aged 15 years and over. This is mainly done through individual face-to-face interviews.

The surveys differ in a number of ways. The HES collects more detailed expenditure data, requiring survey participants to record in a diary all their expenditure over a 2-week period. HILDA focuses more on economic welfare, labour market dynamics, family dynamics and subjective well-being. The HES is a cross-sectional survey that is repeated every 4 or 5 years. On the other hand, HILDA is a household ‘panel’ or ‘longitudinal’ survey. This means that it is also cross-sectional in that it surveys many households at a particular point in time, but it also follows these households across time. This is the first time such a large-scale household panel survey has been undertaken in Australia. However, at this stage, only the first wave of the data has become available so that HILDA is effectively a cross-sectional survey too. The 1993/94 HES covered 8,389 households and nearly 23,000 people with the interviews being equally spread over the period July 1993 to June 1994. The 1998/99 HES was undertaken between July 1998 and June 1999, covering 6,892 households and 13,964 people. The HILDA Survey covered 7,682 households and 13,969 people, with the majority of the Wave 1 data being collected between 24 August 2001 and 21 December 2001.

In using large cross-sectional household data sets to examine the characteristics of financially constrained households we closely follow the approach of several US studies (e.g., Jappelli (1990); Cox and Jappelli (1993); Duca and Rosenthal (1994)). However, these studies focus on households who have been denied access to credit. This kind of micro data is currently unavailable in Australia. Instead, we focus on measures that indicate whether households have difficulty paying their bills, and, by inference, are cash-constrained. In 1998/99 the HES introduced several ways of measuring financial fragility in households. In particular, the survey collected information on whether individuals:

  1. Could not pay their utility bills due to a shortage of money.
  2. Could not pay their registration or insurance on time (rent and mortgage in HILDA).
  3. Pawned or sold something due to a shortage of money.
  4. Went without meals due to a shortage of money.
  5. Were unable to heat their home due to a shortage of money.
  6. Sought assistance from welfare organisations due to a shortage of money.
  7. Sought financial help from friends or family due to a shortage of money.

If a person answers ‘yes’ to any of these questions we define them as having had cash flow problems. Throughout the paper we will interchangeably refer to these people as being ‘cash flow constrained’, ‘cash constrained’, ‘financially constrained’, and ‘financially stressed’.[6] Because the questions on financial stress were only asked in this form in 1998/99 we use econometric techniques to infer the proportion of constrained households in 1993/94 and 2001.[7] However, we defer discussion of the econometrics until we have introduced the data.


Both surveys exclude special dwellings (such as hospitals, institutions, nursing homes, hotels, and hostels) and dwellings in remote and sparsely settled parts of Australia. [5]

However, these concepts are not quite the same as ‘credit constraints’ in the other literature. [6]

HILDA asked very similar questions but differences in sampling technique lead to a large difference in the raw number of households that we would call cash constrained. This is why we use econometric techniques for 2001 as well as 1993/94. However, we are able to use the HILDA answers to verify our results and are thus fairly confident about our findings for 2001. [7]