RDP 8808: Consumption and Permanent Income: The Australian Case Appendix B: Data Sources and Construction

1. Sources of Data

  • National Accounts data – RBA Forecasting Section database (from ABS Quarterly Estimates of National Expenditure)
  • Interest Rate on 90-Day Bank Accepted Bills, All Ordinaries share price index, and M3 – RBA Bulletin database
  • Australian population (end of quarter) – Demography Australia (CBCS, Bulletin No.86) and Population Estimates, Australia (ABS, 3219.0)
  • PAYE income tax paid – Commonwealth Budget Statements (various years)

2. Definition of Variables

  • Consumption – log of real per capita pure consumption of non-durables plus the imputed flow of services from consumption of durables.
  • Consumption expenditure – log of real per capita private consumption expenditure
  • Income – log of real household disposable income
  • Interest rate – log of 90-day bank bill rate, end month of quarter
  • Real interest rate – logged interest rate less the logged inflation rate (defined as 12 m.e. change in the consumption deflator)
  • Share prices – log of real share prices
  • Labour income – the sum of wages, salaries and supplements, plus cash benefits, less PAYE income tax, in real, per capita terms
  • Wealth – real, per capita wealth defined below
  • Population – estimated population, average of end quarter levels

Instrumental Variables (following Hayashi)

  • Real per capita M3 (average of end quarter data, adjusted for new banks)
  • Relative price of imports (deflator for imports of goods divided by consumption deflator)
  • Real per capita government expenditure on goods and services
  • Real exports of goods
  • Real cash benefits to residents

Quarterly consumption and income data are seasonally adjusted, financial data are unadjusted. Except for the real interest rate, all data described as real are deflated by the consumption deflator. Disposable income was used instead of the labour income variable because quarterly data on PAYE income tax could not be obtained.

3. Stock of Consumer Durables

The relevant measure of consumption for the life-cycle hypothesis is a measure which includes the flow of services from the accumulated stock of consumer durables. This is in contrast to the usual National Accounts measure which measures consumer expenditures.

The sixteen expenditure groups in the National Accounts measure of private consumption expenditures can be aggregated into five groups: expenditure on motor vehicles, household durables, food, rent, and other non-durables. The last three categories are generally known as non-durables, and consumption flows are generally equated with expenditures. The first two (motor vehicles and household durables) are known as durables, and yield a flow of services over a considerably longer period than the quarter in which the expenditure is made.

To generate a series for the flow of services from consumer durables, it is first necessary to obtain a series on the stock of consumer durables. There is no officially collected series for either the stock of household durables or the stock of motor vehicles. Accordingly, a series must be constructed using the following identity.

where KDt = stock of durables at end of period t
  CDt = expenditure on durables during period t
  a0 = depreciation rate

and it is assumed that expenditures occur smoothly through each period.

Using estimates for the starting value of the stock of durables and for the depreciation rates, the NIF model database contains estimates of the stock of motor vehicles and household durables from 1959(3). Our own data takes the NIF model depreciation rates (quarterly rates of 6.5 and 5.75 per cent for motor vehicles and household durables, respectively) but assumes different (higher) starting values, especially so for household durables.

Our departure was based on the observation that, taken with the expenditure data, the NIF data implies a rate of increase in the stock of durables at the beginning of the period, especially for household durables, which is significantly higher than for the rest of the period. Accordingly, higher initial starting values were chosen, yielding series for the stock of durable which seem more plausible. These data are available from the authors upon request.

4. Consumption of Durables

Quarterly data for the flow of services from durables are derived from the assumption that this flow is directly proportional to the existing stock of durables in that quarter. The factor relating the flow of services to the stock of durables is determined as follows. In equilibrium, the utility gained from buying and consuming durable goods must be equal to the return available on alternative assets. Given that alternative assets yield a positive rate of return, while durables actually depreciate, the flow of services from durables must equal the sum of the real rate of return and the depreciation rate for durables. Assuming an average real rate of return of 1.25 per cent per quarter, and quarterly depreciation rates of 6.5 per cent for motor vehicles and 5.75 per cent for durables, this yields quarterly service flows of approximately 7.75 per cent and 7.0 per cent of the real value of the stock of motor vehicles and household durables, respectively.[8]

Data for pure consumption are constructed by adding the constructed series for the flow of services from durables to the data for expenditure on other classes of goods.

5. Household Wealth

Estimation of the life-cycle model presented in this paper requires a series for household wealth. Unfortunately, there is no single series for household sector wealth in Australia for the period of this study (i.e. 1959 to the present). Accordingly, a series was constructed using standard splicing techniques from the series of Piggott (1986), Horn (1987), Adams (1987) and Helliwell and Boxall (1978). There are a number of differences in definition between various series, for example the coverage (household or private), and the method of valuation (market value or replacement cost). For the interested reader, further discussions are provided by Piggott and Horn.

Our own series, which is available on request, is based on the following series.

  • Piggott (1986), total personal wealth, Table 5, p.15, plus updates.
  • Horn (1987), net private domestic wealth at market value, last column, Table 2.
  • Adams (1987), market value of total net wealth, Table 4.1, pp.76–7.
  • Helliwell and Boxall (1978), private sector wealth (excluding land) plus value of private land, Table 3, pp.59–60.

It is clear that these series are not totally consistent. However, the different series tend to move quite similarly, so our spliced series should be relatively consistent.


The assumed annual rate of return of 5.0 per cent corresponds quite closely to the estimated discount rate of 4.4 per cent per annum. As an experiment, the estimated discount rate was substituted back into the calculations for the flow of services from durables, and the equations including the discount rate re-estimated. This process quickly iterated to yield parameter estimates which were insignificantly different to those shown in the paper.

Similarly, another cross check on the discount rate (and the series for the stock of durables) is to compare values of the series for the service flow measure of consumption with the expenditure measures. In equilibrium these will be similar, and the mean value of the ratio of the two series will be unity. As expected, the ratio based on our series at times showed significant deviations from unity but the long-run average was very close to unity, suggesting that the constructed series were relatively robust. [8]