RDP 2002-05: Real-Time National Accounts Data 2. Data Issues

2.1 The Availability of Real-time Output Data

Quarterly current price actual output data for Australia have been published by the ABS in timely fashion since late 1960, while corresponding constant price (or, from September quarter 1998 onwards, chain volume) data have been produced since the early 1970s.[6] For the current study we have assembled real-time data on a variety of alternative actual output measures, principally various GDP measures.[7]

To describe in greater detail the extent of real-time data assembled for each of these alternative output measures, it suffices to do so for one of them, say GDP(E).[8] For GDP(E) a total of 109 different vintages of data have been obtained, commencing in 1974:Q4 and running through to 2001:Q4 (see Table 1). The latter 14 of these, which post-date the switch in international standard for national accounting methodology from the United Nations' 1968 System of National Accounts (SNA68) to the System of National Accounts 1993 (SNA93),[9] contain data all the way back to 1959:Q3. Among the remaining 95 vintages of data, however, the situation is more varied, and is summarised in Table 1.

Table 1: Real-time GDP(E) Data
Number of vintages Corresponding quarters Description
29 1974:Q4–1980:Q1, 1985:Q1, 1985:Q3, 1986:Q3, 1987:Q1, 1987:Q3, 1988:Q1 and 1988:Q3 Contain only 14 data points
6 1980:Q2–1981:Q3 Contain data from 1966:Q3 onwards – i.e., between 56 and 61 data points
20 1981:Q4–1984:Q4, 1985:Q2, 1985:Q4, 1986:Q1, 1986:Q2, 1986:Q4, 1987:Q2 and 1987:Q4 Contain data from 1969:Q3 onwards – i.e., between 50 and 74 data points
3 1988:Q2, 1989:Q1 and 1989:Q2 Contain data from 1974:Q3 onwards – i.e., between 56 and 60 data points
37 1988:Q4 and 1989:Q3–1998:Q2 Contain data from 1968:Q1 onwards – i.e., between 84 and 122 data points
14 1998:Q3–2001:Q4 Contain data from 1959:Q3 onwards – i.e., between 157 and 170 data points

As noted above, the position with regard to real-time data availability is similar for the other two main alternative measures of actual output for which a comparable number of vintages are available, namely GDP(I) and GNFP (see also Table 2).

Table 2: Measures of GDP
Timeline for production of quarterly GDP measures
1960 Publication of current price GDP(E) and GDP(I) estimates commences in ABS Cat No 5206.0 (National Income and Expenditure)
1971 Constant price GDP(I) estimates first included in ABS Cat No 5206.0
1974 Constant price GDP(E) estimates first included in ABS Cat No 5206.0
1988 Publication of GDP(P) estimates commences in ABS Cat No 5222.0 (Gross Product, Employment and Hours Worked)
1988–1991 Gradual move to concurrent publication of ABS Cat Nos 5206.0 and 5222.0, allowing simultaneous production of GDP(E), GDP(I) and GDP(P) measures
1991 Introduction of concurrent publication of the average GDP measure, GDP(A), together with GDP(E), GDP(I) and GDP(P)
1998 Shift by the ABS to chain volume rather than constant price estimates in the national accounts, and introduction of an official chain-volume GDP series
Notes: References to constant price GDP estimates, for vintages prior to the introduction of chain-linking in 1998, denote figures prepared using average prices in a given base year. This base year was typically updated roughly every 5 years. Constant price GDP(I) was estimated by summing the deflated components of GDP(E) and then adding the statistical discrepancy deflated by the GDP(E) deflator.

2.2 Which Output Measure to Focus On?

The availability of real-time data for various alternative measures of actual output raises the question: what is the most appropriate measure upon which to focus? We select two series on which to concentrate, expenditure-based GDP and a hybrid GDP measure spliced together from a succession of different GDP series. By way of background to these choices, Table 2 provides a brief history of the production of quarterly GDP statistics by the ABS since these were first published in December 1960.

In deciding which domestic output series to study, we are primarily interested here in whichever series was regarded by analysts at the time as representing the best available output indicator. Prior to 1991 this appears to have been GDP(I), which was regarded as a more reliable measure of the true state of the economy than GDP(E). As Budget Statement 2, Budget Paper No 1 of the 1980–81 Budget put it:

In Australia, the income-based estimate of GDP has traditionally been used as the basic measure of product growth, with the statistical discrepancy being assigned to the expenditure side of the accounts.

With the introduction of a GDP(P) measure in 1988 it became possible to construct an average GDP measure, GDP(A), from the three alternative income, expenditure and production-based GDP estimates. An ABS analysis, released with the June quarter 1990 national accounts, concluded that:

All in all, the facts seem to point to GDP(A) as being the best indicator of short-term movements [of gross domestic product]. (Aspden 1990, p 65)

However, at this time GDP(P) was still only being published with a lag relative to GDP(E) and GDP(I), preventing sufficiently timely construction of GDP(A) for it to become the preferred real-time focus of policy-makers. Not until 1991, when concurrent publication of all four GDP measures (E, I, P and A) was achieved, do we therefore regard analysts as switching their primary interest to GDP(A).

Finally, GDP(A) then appears to have remained the principal output measure for analysts until the shift from constant price to chain-volume estimates in the national accounts in September quarter 1998. This shift saw the simultaneous abolition of the GDP(A) terminology and its replacement by an official chain-volume GDP series as the new main focus for analysts.

This history then explains the latter of our two choices of real-time GDP series upon which to concentrate in Section 3, namely the hybrid of GDP(I) until September quarter 1991, GDP(A) from December quarter 1991 to June quarter 1998, and chain-volume GDP thereafter. As noted above, we have chosen to supplement examination of this hybrid series with study also of the real-time behaviour of GDP(E) estimates. The reason for this choice is that, although the hybrid series just described probably represents the main focus of analysts throughout the past three decades among alternative output measures, the bulk of output forecasting conducted by the RBA and Treasury over this period has been expenditure-based, reconstructing aggregate output from forecasts of expenditure components in the national accounts such as household consumption, business investment and so forth. Given that policy decisions are likely also to have been influenced, at any particular moment, by forecasts of future output, it is therefore also of interest to have a feel for how significant the real-time problem has been for GDP(E), since errors in the measurement of actual GDP(E) would likely have flowed through to errors in forecasts of this quantity.


Strictly speaking, these early GDP estimates were produced by the forerunner to the ABS, the Commonwealth Bureau of Census and Statistics (CBCS). The CBCS started publishing annual constant price GDP estimates for Australia in 1963, but the first quarterly (income-based) estimates published were for the September quarter 1971. [6]

Precisely, the series collected are: expenditure and income-based gross domestic product (GDP(E) and GDP(I)) and gross non-farm product (GNFP); production-based and average gross domestic product (GDP(P) and GDP(A)) from the late 1980s onwards; and finally, since the shift to chain-volume gross domestic product measures from September quarter 1998 onwards, the official ABS gross domestic product series (GDP). Data have been obtained on both a current and constant price basis, and in both original and seasonally adjusted terms. [7]

The situation is similar for GDP(I) and GNFP, while for GDP(P) and GDP(A) we have assembled real-time databases of all vintages of these series since their first production in the late 1980s. Likewise, a complete real-time database has been assembled of all vintages of the official ABS chain-volume GDP series, which has been produced since September quarter 1998. [8]

SNA 1993 was prepared and published jointly by the Commission of the European Communities (Eurostat), the IMF, the OECD, the United Nations and the World Bank. [9]