Growth accounting: further details and robustness checks
As we have seen, we necessarily make a number of assumptions when implementing the growth accounting exercise. How robust are our findings to key assumptions? This is shown in the rest of the table, where for ease of reading we just show the results for this century. All results for all other periods are available.
The first row in Panel 2, Table 5, shows the results when only software is included as an intangible. Thus this row corresponds closely to current National Accounts practice, although copyrights and mineral exploration are also capitalized in official data. As can be seen, relative to the very top panel, which excludes software, capitalization of software raises ln V/H and lowers, very slightly, ln TFP. Note from column 5 that the contribution of software is 0.09%pa, against the total intangible contribution of 0.51%pa.
The next row capitalizes both software and R&D and thus a comparison with the software line estimates the difference due to R&D capitalization (to be implemented in the UK by 2014).
Relative to software, the contribution of intangibles rises very slightly and ln TFP falls very slightly. So capitalization of R&D adds about 0.03%pa to input contribution and TFP falls by the same.
15
The next two rows halve and double the assumed intangible depreciation rates. This raises and lowers the contribution of intangible capital respectively, as would be expected. They more or less directly affect ln TFP, so that, if for example, intangibles depreciated half as fast as we have assumed, ln TFP falls from 0.90%pa to 0.80%pa.
Finally, since own account organizational capital is particularly uncertain, the final row reduces such spending by 75% (that is, managers are assumed to spend 5% of their time on organizational capital). In this case, the contribution of intangible capital falls from 0.51%pa to 0.45%pa and ln TFP rises from 0.90%pa to 0.93% pa.
One way of looking at the robustness of these results is to calculate the fraction of overall ln V/H accounted for by intangibles, ln TFP and ln L/H under the various different scenarios. It is in fact quite robust. As row 3, top panel shows, without intangibles, the ln TFP fraction is 0.53 and ( ln TFP + ln L/H) is 0.62, a result that is very similar with just software or just software and R&D. With intangibles, the fractions are 0.40 for ln TFP, 0.63 for ln TFP + ln K/H(intan) and 0.70 for ln TFP + ln K/H(intan) + ln L/H. But the interesting thing to note is that these fractions are almost identical with the experiments on depreciation and organizational capital. Thus the inclusion of the full range of intangibles lowers the share of the contribution of ln TFP, but consistently raises the share of the contribution of ln TFP, intangible capital deepening and labour composition combined, such that the latter has accounted for 70% of ln V/H over this century.