RDP 2002-05: Real-Time National Accounts Data Appendix B: An Alternative Approach to Assessing the Frequency and Persistence of Alterations to Output Growth Estimates

In Section 3.2 we examined the frequency and persistence of initial mismeasurement of output growth by comparing the initial estimates of quarterly or four-quarter-ended growth for each quarter in history with the estimate for that quarter contained in the latest available (2001:Q4) chain-volume GDP data. This assumes that this latest data vintage provides the best available measure of the ‘true’ growth of real output in each quarter over history. For periods far back in time, however, a technical issue arises as to whether this is an appropriate assumption.

This relates to the change made in 1998 to the national accounting system under which subsequent accounts have been prepared, and the difficulty of accurately adjusting estimates of real output for periods back in the 1960s, 1970s and 1980s to reflect these changes (so as to give a time series on a consistent accounting basis). To avoid this problem one could instead use the last set of accounts prepared under the old system of national accounting (SNA68), namely those from June quarter 1998, as the best available measure of the ‘true’ growth of output over history. However, this would raise a separate issue relating to the appropriateness of the price data used in those accounts for obtaining estimates of real output growth for quarters far back in time.

This latter issue relates to the fact that the constant price GDP estimates contained in the June quarter 1998 accounts were only directly computed using average 1989/90 prices (the then price base year) for the most recent 5–10 years. Prior to this, estimates based on earlier base years were then spliced on at 5–10 year intervals to give a continuous historical time series, with this splicing done at a disaggregated level (and therefore possibly affecting growth rates at the aggregate GDP level). Growth rates based on these estimates, for periods prior to the late 1980s, may thus possibly be a less accurate guide to the ‘true’ real growth occurring in the economy at that time than earlier constant price estimates prepared using the appropriate earlier price base year.

To avoid altogether these various problems associated with having to select a recent data vintage to represent the ‘true’ growth of output over time, a different approach would be to assess the frequency and persistence of mismeasurement of output growth by comparing initial growth estimates in each quarter with those made (say) one year, three years and five years after the event.[22] The latter horizon broadly reflects the period over which formal revisions are typically made to GDP estimates – by which we mean revisions relating to the incorporation of improved underlying data (for example, resulting from new information about the timing of particular items of activity, from more accurate taxation data which only becomes available with a lag of several years, or from updated census data with which to adjust relevant GDP components, available at five-year intervals).

To determine whether this alternative approach to assessing the scale and persistence of changes to output growth estimates alters the broad picture described in Section 3.2, Figure B1 shows the extent to which the initial estimates of four-quarter-ended growth in our hybrid GDP measure have, over history, been amended one year, three years, and five years after the fact.

Figure B1: Errors in Four-quarter-ended GDP Growth
Alternative methodology
Figure B1: Errors in Four-quarter-ended GDP Growth

With regard to the persistence of initial mismeasurement, changes to the assessment of four-quarter-ended growth made in the first year following the period under consideration have, on average, been somewhat larger than those made over the next two years. These, in turn, have typically been somewhat greater than those made between three and five years after the event. The differences, however, are not dramatic.

Overall, the results of this alternative analysis confirm the general conclusions found earlier. There appears to be sufficient uncertainty surrounding initial estimates of GDP growth that analysts need to keep this in mind when treating such estimates as a barometer of the current state of the economy. At the same time, there does not appear to be any bias in these estimates, and there is evidence that, with the improvements in national accounting methodology and data collection/reconciliation over the past three decades, the scale of the real-time problem may be smaller now than it has been in the past.


Note that this alternative approach does not aim to abstract completely from adjustments related to constant price base year changes, when examining the alterations to output growth estimates for quarters far back in history. For example, consider estimates of four-quarter-ended growth to the March quarter 1985. For our hybrid GDP measure the initial estimate for this quarter is of four-quarter-ended growth in GDP(I), measured on a constant price basis using average 1979/80 prices. Three years later, in the March quarter 1988 accounts, the estimates of growth for this period shift to using average 1984/85 prices, which we would expect would give a more accurate indication of actual real growth in the year to March quarter 1985 than estimates based on 1979/80 prices. In assessing the scale of the real-time problem facing analysts in early 1985, we thus do not wish to abstract from the change made in the March quarter 1988 accounts to the constant price base year being used. Rather, we only wish to abstract from subsequent further price base year changes (e.g. to average 1989/90 prices in the December quarter 1992 accounts), which might arguably have made the estimates for output growth in the year to March quarter 1985 less, rather than more, accurate. [22]