RDP 9107: The Cost of Equity Capital in Australia: What can we Learn from International Equity Returns? 4. Measures of the Cost of Equity

(a) Data on Average Realised Returns

Typically, researchers look at realised returns in a country by looking at excess returns for investments denominated in that country's currency. However, in an integrated world capital market all investors have access to all markets, and it seems appropriate to look at all excess returns in a single currency, such as the SDR. This becomes more important in a world where exchange rate movements can be large, and might significantly change the picture given by looking at each country in its home currency.

Table 2 presents some estimates of the equity premium for 11 countries and the world index over a 21-year period. Two measures are shown. First is a measure of the home currency equity premium over this period, derived by taking the average local currency stockmarket return for each country (from the geometric rate of return) and subtracting the average short-term interest rate over the same period.[12] Second, I calculate an SDR-denominated equity premium to a hypothetical world investor. This is derived by taking the average home currency return, then adding a factor for the average change in the SDR exchange rate over the period, and then subtracting an average world interest rate.[13]

Table 2: Estimates of the Equity Premium, Dec 1969–Dec 199
Country Local Currency Returns Short-term Interest Rate Equity Premium in Local Currency Average Exchange Rate Change Equity Premium to World Investor
Australia 9.5 11.7 −1.9 −3.4 −1.9
Canada 10.5 9.3 1.1 −2.0 0.4
France 11.9 9.6 2.1 −1.3 2.5
Germany 7.5 6.8 0.6 2.7 2.3
Italy 9.4 12.5 −2.7 −4.4 −3.0
Japan 13.8 6.7 6.6 3.0 8.6
Netherlands 11.1 6.9 4.0 2.0 5.1
Sweden 15.9 9.2 6.2 −2.1 5.3
Switzerland 5.3 3.9 1.3 4.2 1.7
UK 14.7 10.3 4.0 −2.7 3.5
US 10.0 7.5 2.3 −1.7 0.3
World 9.8 7.9 1.8 0.0 1.8

The calculations indicate that the returns from investing in the Australian equity market over this 21-year period would not have exceeded the returns from investing in short term assets over the period, either for an Australian investor, or a world investor. That is, the equity “premium” in Australia was actually negative in this period. Most other countries show positive equity premia in SDR terms: my measure of the world equity premium over this period is 1.8 per cent.[14] The highest yielding stockmarket on this measure is Japan which shows an equity premium to local investors of 6.6 per cent, and one of 8.6 per cent to a world investor, reflecting the rise of the yen over this period.

The finding that Australia shows a negative equity premium over this period deserves further comment. If we used the Statex accumulation index (which is available only from December 1971) we would get a slightly higher growth rate for equities in the period for which both indices are available, though the overall conclusions would still be very similar. (Any differences may reflect a “small firm effect” since the Statex index has broader coverage than the MSCI index: however, if other countries have similar effects then this bias will exist there as well.) Alternatively, if we deflated the MSCI data by the average growth in the CPI since December 1969, real equity prices would show a very small increase over the period, though still significantly less than the real increases that other countries would show. This suggests that the weak performance of the Australian market is a fairly robust finding, at least for this period. Part of the reason for this weakness may be that the starting point was near a major peak in the Australian market associated with a metals boom. Any starting point will be somewhat arbitrary: other starting dates would give different, and sometimes higher, returns for the Australian market.

If a 21-year period is sufficiently long for errors in expectations to average to zero, the average realised returns in Table 2 provide measures of the required rate of return in each country. According to this rational expectations view, because Australian stocks have shown relatively low returns, they must have some other desirable properties that allow them to have lower required rates of returns than other countries. In addition, Japanese equities which have had high rates of return must have some properties which make them unattractive to investors, and which increase their required rates of return. It hardly need be said that this implication is contrary to the conventional wisdom of recent years.

However, it is not obvious that realised returns will be a good proxy for expected returns. A fundamental reason is that realised returns will also reflect unexpected developments that occurred during the period. In the case of Australia, two factors spring to mind. First, the growth of the industrial sector will be significantly affected by GDP growth. However, per capita GDP growth in the period 1970–1990 averaged around 1.5 per cent per annum in Australia, compared with around 3.7 per cent for Japan.[15] Of course, according to the rational expectations hypothesis, this difference in growth rates will not have mattered for realised returns unless it were unexpected. If it was not fully expected, as seems reasonable, Japanese equity prices would have increased relative to Australian equities as it became apparent, boosting returns in Japan. That is, relative GDP growth performances can probably explain ex post outcomes to a large extent.

Second, the performance of the resources sector (and indirectly the industrial sector) will be significantly affected by metals prices. Between December 1969 and December 1990, real metals prices fell by a massive 63 per cent.[16] It seems most unlikely that this fall (equivalent to nearly 5 per cent per year) could have been fully anticipated, so it probably also helps explain the poor outcome in Australia.

Problems with the realised returns approach will not necessarily go away by simply using longer periods of data. Data on long term excess returns are only available for a few countries: the US, Australia and the UK. The data that do exist suggest that the equity premium is something over 6 per cent for all three countries. However, as Poterba (1991, pp. 24–25) points out, the standard deviation of annual excess returns data is so large as to prevent any convincing conclusions about the cost of equity. In addition, there are at least three reasons for thinking that there are limits to what we can learn about the cost of equity from standard measures of the historical equity premium in Australia.

First, the per capita growth performance of the Australian economy for the century as a whole has been (with the UK) the lowest in the OECD. Cumulative growth in GDP over the period 1900–1987 was around 80 per cent higher in the total OECD than in Australia. Unless Australia's rather dramatic slide down the OECD rankings was anticipated, and hence already reflected in stock prices, realised returns in Australia will be biased downwards as a measure of expected returns. That is, to rely on realised rates of return, we have to try to adjust for unexpected outcomes.

Second, historical estimates of the equity premium in Australia (e.g. Officer (1989)) tend to rely on data from Lamberton (1958 a,b) for the period to 1955. However, Lamberton's data contain few resource stocks. Yet the resources sector is generally thought to be riskier than the industrial sector, so exclusion of the resources sector may understate the true required rate of return for the economy as a whole. That sector currently accounts for around 35 per cent of total Australian market capitalisation, so one must ask whether the historical data are really representative of the Australian market today. In any case, we must also ask if other fundamental structural changes have occurred in the economy, making it dangerous to infer too much about the present, from data from the first half of this century.

Third, if countries have had permanently different debt-equity ratios, this will have affected returns. Australian equities have, at least in recent years, had lower debt ratios than other countries, so in this period they should have been less risky and should have had lower rates of return. Adjusting for this is difficult: while data for stockmarket returns over long periods of time are not good, data on debt/equity ratios would presumably be far worse. All that can be done is to highlight the possibility that differences in leverage across countries might overturn any inferences on the cost of equity that are based purely on historical equity premia.

(b) Data on E/P Ratios

Table 3 provides some data on E/P ratios. The first column contains average E/P ratios for five countries and the world index for the period 1984–1990 from the MSCI database. The second column makes two adjustments to the data. First, it adjusts for the fact that the data use retrospective rather than prospective earnings. For this adjustment, the average annual growth in nominal GDP is used to proxy nominal earnings growth in each country, along the lines of equation 3 in Section 3(b). Second, the arbitrary assumption that real earnings were expected to grow by 2 per cent per annum in each country is then made, to derive simple proxies for the real cost of equity.[17] The last column provides estimates of the real cost of equity in 1988 from McCauley and Zimmer (1989) and the Australian Manufacturing Council (1990).

Table 3: Data for E/P Ratios
Country Average
E0/P1 Ratio
(1984–1990)
E1/P1 Ratio
Plus Growth
Factor
Real Cost of
Equity, 1988
(per cent)[18]
Australia 0.085 0.114 13.4
Germany 0.070 0.094 3.0
Japan 0.027 0.048 5.0
UK 0.086 0.113 8.0
US 0.077 0.103 11.0
World 0.062 0.086 n.a.

The estimates from McCauley and Zimmer, and the AMC Report have corrected measured E/P ratios for cross country differences in a number of factors which affect measured earnings. These include adjustments for the effect of inflation upon depreciation allowances, on inventory profits, on nominal interest payments, and for the effect of crossholdings between Japanese companies. These estimates suggest that in 1988 the cost of equity was higher in Australia than overseas. Given the volatility of E/P ratios, it would be interesting to see the comparison over several years, so as to abstract from temporary factors. However, my own simple measure derived from the MSCI data averaged over 7 years is also consistent with the hypothesis that the real cost of equity in Australia is a little higher than some other countries. It should be noted that these estimates are for the cost of equity, given existing debt/equity ratios. But the Australian market has had lower debt ratios over this period: if Australian firms had ratios similar to foreign countries one would expect that earnings/price ratios would be even higher in Australia.

A rough check of the plausibility of the framework (including the assertion that E/P ratios proxy the real rather than nominal cost of equity) is provided by the implied cost of equity for the world market as a whole. If we start with the estimate for the real cost of equity of 8.6 per cent for the world market as a whole, and then assume an average real short-term interest rate of 4 per cent,[19] we obtain a world equity premium of around 4½ per cent, which is somewhere between standard historical estimates and the values that are suggested by theoretical models of asset pricing.

Irvine (1991) discusses a number of drawbacks in the use of E/P ratios as measures of the cost of equity. Two major criticisms are discussed below.

First, E/P ratios are volatile and have a cyclical pattern, since stock prices fall before earnings as the economy goes into a downturn, and rise before earnings during the recovery. It may be that Irvine's view on the volatility of E/P ratios is due to his incorrect use of E/P ratios as nominal discount rates. Thus, he apparently subtracts a nominal interest rate from the E/P ratio in an attempt to obtain a real equity premium for the UK (p. 15). However, as discussed in Section 3(b), E/P ratios are measures of the real cost of equity, so it is not surprising that he obtains a number that is volatile, and sometimes negative. As to the cyclically of E/P ratios, some of this will be due to the tendency of markets to use backward looking E/P ratios: consensus forecasts of prospective earnings would generally yield a smoother ratio. Alternatively, if E/P ratios are averaged over periods of time which include both slowdowns and upturns, the effect of this cyclically will be removed. Thus, comparisons across countries should use recent data averaged over a number of years.

Second, despite adjustments to accounting earnings to make them closer to true economic earnings, Irvine claims that these measures will be poor indicators of free cash flow, which he describes as the real source of shareholder returns. In addition, differences in accounting rules mean that accounting earnings will not be comparable across countries. These points have been widely canvassed in the discussion over whether the cost of equity is lower in Japan than in the US. A number of points emerge from the most recent literature. First, adjusting Japanese earnings to a similar basis to US depreciation rules increases E/P ratios in Japan, but leaves a significant gap still unexplained. In any case, price/cash earnings ratios tend to show differences that are nearly as large as with conventional P/E ratios, so the treatment of depreciation may not be that important.[20] Second, adjusting for intercorporate holdings which are very significant in Japan does increase E/P ratios, but making this adjustment will also increase US E/P ratios, so Japanese ratios remain significantly below US ratios.[21] Third, correcting Japanese earnings for unrealised capital gains on land does increase E/P ratios significantly, and may account for the difference in E/P ratios.[22] That is, a number of adjustments to E/P ratios have been tried, but only a phenomenon as extreme as the recent massive land price inflation in Japan appears able to significantly change the picture given by standard E/P ratios. In any case, as McCauley and Zimmer (1989, Table 1) show, some of the adjustments that can be made to E/P ratios act in different directions in different years, hence using averages over a number of years may remove many of the problems. Thus, one conclusion from this body of literature might be that E/P-based rankings of the cost of equity are not easily overturned. It would still be interesting, however, to see how Australian E/P ratios are affected by this barrage of adjustments.

It is apparent, therefore, that the earnings yield approach is not without flaws, but given that the required rate of return on equity is unobservable, we must expect that any proxy for it will have weaknesses. The main strength of E/P ratios is that they do contain information as to the market's valuation of earnings flows, so it would seem dangerous to ignore them.

From the point of view of the firm, E/P ratios may provide a useful signal of when equity issues will be easiest, and when averaged over the recent past, they may provide a reasonably good indicator of the cost of equity that firms can use for investment appraisal. In addition, if national E/P ratios are averaged over time, they contain valuable information about the way that the world market values the earnings of Australian companies relative to the earnings of foreign companies.

Footnotes

Following Mehra and Prescott (1985), Ibbotson Associates (1990), and the CAPM literature, the equity premium is calculated using short-term interest rates rather than long-term (risky) ones. All interest rate data are from OECD Main Economic Indicators and the IMF International Financial Statistics. I thank Mark Rider and Michele Bullock for providing the data. For the world interest rate, I use an average of interest rates in the US, Japan, Germany, France, and the UK, using the currency weights of the SDR. The interest rates used for each country are slightly different, but an attempt was made to get a measure as close as possible to a three-month government security for each country. Where the rate is a money market or interbank rate there will be slight biases, but these will be fairly minor compared with the differences from other sources. [12]

For most countries the equity premium to the hypothetical world investor is similar to the conventional local currency equity premium. This reflects the fact that interest rate differentials in this period have largely been offset by exchange rate changes. Indeed, it is straightforward to show that if exchange rates are determined in an interest parity framework, the two definitions of the equity premium are exactly the same. [13]

This is somewhat below the 6 or more per cent which various studies have found for longer time periods in a few different countries. This is largely attributable to the poor performance of most stockmarkets in the first half of the 1970s. It may be, however, that the required equity premium is now lower than earlier data suggest, as there is no reason why required rates of return might not have fallen over time. If so, this would be consistent with the literature, starting with Mehra and Prescott (1985), that is unable to explain the magnitude of historical equity premia by theoretical models of asset pricing. [14]

The data for GDP growth are taken from Table B3 in Maddison (1989), updated to June quarter 1990 from OECD sources. The population adjustments are approximate, and use the average population rate growth rates from Table 1.2 for the period 1950–1987. [15]

Calculated using the IMF metals index in SDRs, deflated by the CPI of the industrial countries, using data from International Financial Statistics. [16]

The assumption of a 2 per cent annual growth rate of earnings should not be considered to apply literally for every year to every stock in every country. Start-up companies will obviously have very high (and volatile) expected growth rates, while mature companies may even have negative ones. However, when averaged over companies, and over several years the assumption of similar growth rates may be reasonable. Alternatively, one could use forecasts of earnings growth in each country, but typically these would show little variation in a group of countries of similar stages of industrialisation. The one component of growth that might be forecastable is the part driven by the population growth of a company's dominant (usually home) market. Australia's population growth is expected to remain higher than other countries: on this view, a higher growth rate could be applied to some Australian companies. [17]

Source: Australian Manufacturing Council (1990, p. 89). [18]

Bullock and Rider (1991) obtain real interest rates around 4 per cent for most countries in their sample. [19]

French and Poterba (1990, pp. 17–18). [20]

French and Poterba (1990, p. 18), Ando and Auerbach (1990, p. 12). [21]

Ando and Auerbach (1990, p. 18). However, this adjustment has only been made in Japan, and could increase E/P ratios (to a lesser extent) in other countries which have also had land price appreciations. [22]