RDP 9107: The Cost of Equity Capital in Australia: What can we Learn from International Equity Returns? 6. Can Stockmarket Risk be Explained by Real Earnings Risk in Australia?

Section 5(a) suggested that Australian equity returns may be relatively risky when compared with the world market, and after leverage considerations are taken into account. This section looks closer at whether there are fundamental factors which can explain the local stockmarket risk. That is, while the previous section looked at risk in financial markets, this section will analyse data from the real economy. If there is no evidence of this fundamental risk, we may be forced to conclude that the financial market risk is due to some irrational factor. An example might be simple excess volatility to fundamentals, as Shiller (1981) and others have suggested for the US stockmarket.

This suggests that we should look at actual dividends or earnings across countries over a long period of time, and to see whether earnings in particular countries have traditionally showed excess sensitivity to world movements. Good data on dividends are available across countries, but countries may have different norms of dividend distribution. In particular, there may be differing propensities to increase or cut dividends in response to changes in earnings. Thus, earnings data may be a theoretically better way to address the question.

However, data on reported earnings are not available on a consistent basis because of differences in accounting rules. A compromise is to use data from national accounts sources which are usually constructed on quite similar bases. Accordingly, OECD data are used for total operating surplus, net of depreciation, in each country. An advantage of these data is that they measure the total return to capital, before payments to debt and equity: that is, they ignore the financing mix and measure true economic earnings flows.

There are, of course, some weaknesses with these data: they include a number of components that are not normally included in financial-market research. As well as returns to private corporate trading enterprises, they include returns to unincorporated enterprises, dwellings, public traded enterprises and general government. However, there are arguments for including all of these: investment in dwellings can be regarded as an alternative to investing in equities, and must be priced using similar principles. Similarly, to the extent that public ownership can be treated as a veil for individual ownership it should be considered as normal equity. In addition, the data will remove any differences across countries according to whether some sectors such as utilities are government-owned or privately-owned.

The methodology is to derive a measure for real operating surplus in each of 11 countries, and then to aggregate them to form an index for the total.[33] Table 7 shows the regression results from estimating the extent to which earnings in each country “respond” to world earnings.[34] The first estimates use annual data from 1961 to 1988, and regress the change in earnings in each country against a constant and the change in world earnings. For all countries except Sweden (where the results appear to be driven by a single outlier, 1977) there is significant explanatory power, suggesting that world factors do systematically affect earnings in different countries. For Australia, the standard error on the estimate is 0.29, so there is evidence that Australian operating surplus responds more than one-for-one with world operating surplus.[35]

Table 7: Correlating National Earnings with World Earnings
Country const World
Operating
Surplus
(t)
adjR2 const World
Operating
Surplus
(t−1,t,t+1)
adjR2
Australia −1.0 1.57 0.52 0.2 1.17 0.51
Canada 2.1 0.83 0.28 3.4 0.35 0.33
France 0.3 0.75 0.43 −0.7 1.00 0.53
Germany 0.0 0.77 0.48 0.0 0.79 0.47
Italy 2.4 0.37 0.08 2.7 0.24 0.46
Japan 2.4 0.70 0.18 1.6 0.92 0.19
Netherlands 0.4 0.65 0.22 1.5 0.29 0.21
Sweden 1.8 −0.19 −0.03 2.1 −0.37 −0.03
Switzerland 0.5 0.35 0.13 −0.4 0.61 0.37
UK −1.8 1.44 0.40 −2.8 1.93 0.45
US −1.0 1.36 0.76 −0.6 1.20 0.79

However, because some countries may lead or lag the world business cycle, a one-period lead and lag are also included as explanatory variables. These results are shown for the period 1962–1987. The coefficients on the three explanators have been summed to give the total effect. Standard errors are not provided: it will suffice to note that they are large, with only 22 degrees of freedom. However, these estimates are also consistent with the hypothesis that Australian earnings show more cyclical volatility than earnings in most other countries. That is, there appears to be some evidence from the real sector to lend support to the evidence on asset betas that was presented in Section 5(a).

Footnotes

Overseas data are taken from OECD National Accounts: Main Aggregates, Vol. 1. Australian data are taken from ABS Australian National Accounts. To aggregate, I calculate the change in real earnings in each country, and then weight to give the aggregate change in world earnings. Three alternative sets of weights were used, but the results are not especially different, so only one set of results is quoted. [33]

The analysis would ideally use rates of return rather than changes in earnings. Consistent estimates of the capital stock for each country were not available, hence it is not possible to get true rates of return. However, rates of return will be dominated by movements in real earnings since earnings show significant volatility, but capital stocks will change only slowly. [34]

The change in real commodity prices was also included as a regressor for each country, but for most countries, including Australia, there was little extra explanatory power. One might have expected part of Australia's excess cyclicalily to explained by real commodity price movements, which might also be correlated with world earnings. However, there was only a small correlation between world operating surplus and commodity prices, hence multicollinearity is not the explanation, at least in this data set. Another data set with more observations might yield different conclusions. [35]