RDP 2002-05: Real-Time National Accounts Data Appendix A: Revisions to Expenditure-based GDP Estimates

Andrew Stone and Sharon Wardrop

September 2002

This appendix replicates, for GDP(E), the analysis of our hybrid measure of GDP carried out in Section 3. In addition to switching our focus to GDP(E), one further change has been made in the treatment below, relative to that in Section 3. This is that, in this appendix, we use the June quarter 1998 national accounts data as giving our best measure of the true growth rates of GDP(E) over history, rather than the December quarter 2001 vintage of data used earlier for our analysis of hybrid GDP. We choose this quarter as it is the last vintage of GDP(E) data produced under the old SNA68 system of national accounting, prior to the changeover to SNA93. Using this vintage of GDP(E) data to represent the ‘true’ measure therefore incorporates the maximum possible set of revisions to each earlier quarter's estimates while still ensuring that comparisons between this and earlier data vintages involve a comparison of ‘like with like’.[21]

As Figures A1 through A5 illustrate, the results for the analysis using GDP(E) are similar to those obtained for our hybrid measure of GDP. This being the case, our general conclusions remain unchanged. Indeed, simple comparison of Figures A1 to A5 with Figures 1 to 5 confirms that the real-time measurement problem is, in fact, somewhat greater for GDP(E) than for hybrid GDP, consistent with the preference of analysts and the ABS for first GDP(I), and then GDP(A), as more reliable measures of output than GDP(E).

Figure A1: Changing Estimates of Quarterly GDP(E) Growth
Figure A1: Changing Estimates of Quarterly GDP(E) Growth

Note: This figure mirrors the analysis in Figure 1 in the main body of the paper.

Figure A2: Changing Estimates of Four-quarter-ended GDP(E) Growth
Figure A2: Changing Estimates of Four-quarter-ended GDP(E) Growth

Note: This figure mirrors the analysis in Figure 2 in the main body of the paper.

Figure A3: Errors in the Contemporaneous Measure of GDP(E)
Quarterly growth rates
Figure A3: Errors in the Contemporaneous Measure of GDP(E)

Note: This figure mirrors the analysis in Figure 3 in the main body of the paper.

Figure A4: Errors in the Contemporaneous Measure of GDP(E)
Four-quarter-ended growth rates
Figure A4: Errors in the Contemporaneous Measure of GDP(E)

Note: This figure mirrors the analysis in Figure 4 in the main body of the paper.

Figure A5: Errors in Four-quarter-ended GDP(E) Growth
As measured at various time horizons
Figure A5: Errors in Four-quarter-ended GDP(E) Growth

Note: This figure mirrors the analysis in Figure 5 in the main body of the paper.

Note that for Figures A1 and A2 we have chosen to display the successive vintages of data for different quarters from those used in Figures 1 and 2. The quarters selected, June quarter 1986 and December quarter 1990, have been chosen to illustrate the scale of changes to quarterly and four-quarter-ended GDP(E) growth rates not infrequently encountered over time. As Figures A3 and A4 demonstrate, quarters displaying more extreme changes than these two could once again have been selected.

Footnote

An alternative would be to use the GDP(E) series from the December quarter 2001 national accounts as giving the ‘true’ measure of GDP(E) growth. This would have the advantage of allowing for an additional 14 vintages worth of data revisions, but at the cost of no longer having all the GDP(E) vintages under comparison having been prepared under a common system of national accounting (see also the discussion in Appendix B). It transpires that adopting this alternative choice makes essentially no difference to the results described hereafter. [21]