RDP 9302: A Decade of Australian Banking Risk: Evidence from Share Prices 5. Results

The effect of the share market crash in October 1987 dominates the results and swamps all other movements in our derived series.^[15] This effect was also observed in work done with US data (Levonian (1991b) and Schellhorn and Spellman (1991)). The graphical results are presented omitting the last six months of 1987 so that other variation may be discerned. Although the contingent claim model assumes that the stock market is always efficient, there are grounds for considering the crash to be an anomalous event that should be excluded from the analysis.

5.1 Capital

Looking first at the banking sector as a whole, weighted averages are constructed to combine the individual bank results at each time period, with each bank weighted by its share of the total market value of assets in the sample. Graph 1, presents the weighted average market capital ratio for each case. The estimated capital ratio rose from below zero in early 1983 to a peak of about 2.5 percent in late 1988. The capital ratio declined a little until early 1991 and, since then, recovered to about two percent as of March 1992.

Because of the use of the weighted average, the secular pattern of the capital-asset ratio is most heavily influenced by the three largest banks in the sample. Individual bank results vary considerably, with some banks' capital-asset ratio rising greatly over the 1980s, and others remaining static or falling slightly.

The market capital-asset series varies considerably over time. To determine the source of this variation the change in the market capital asset-ratio was decomposed into movements attributable to changes in equity, licence value and the contingent deposit guarantee liability, each relative to assets.^[16] The results showed that movements in the market capital-asset ratio were dominated by fluctuations in the ratio of shareholder equity to assets, with about 95 percent of movement in the capital ratio attributable to movements in market equity. Changes in the ratio of aggregate market equity to assets was then broken up into components attributable to changes in the number of shares on issue, changes in the share price and changes in aggregate bank assets. This decomposition showed that most of the movement in the equity to asset ratio was due to movements in share prices, reflecting market revaluation of banks' capital, rather than active changes in bank capital through share issuance.

However, a number of banks, including the three major private banks, have made some large capital raisings, particularly in the latter half of the decade. Some of the increase in bank capital may have been brought about by various regulatory requirements introduced through the 1980s. Prior to the adoption of the BIS risk-based capital adequacy requirements in 1988 established banks were required to hold capital equal to at least 6.5 percent of total assets, while new entrants, such as building societies converting to bank status and foreign banks taking up a banking licence in Australia, were required to observe a higher capital-asset ratio. In practice, new entrants tended to hold capital well above the required minimum. Hence, the entry of newer banks into our sample may account for some of the increase in the weighted average market capital-asset ratio since 1985. Moreover, while most banks were well placed to comply with the BIS risk-based standards, the introduction of those standards may have led some banks to issue more equity. It is likely that some portion of the measured variation in capital ratios is due to these regulatory changes.

5.2 Asset Volatility

Graph 2, shows the weighted average asset volatility across all banks in our sample. Asset volatility displays a slight upward trend over the 1980s. This may simply reflect the entry of new, smaller, banks into our sample. The most noticeable feature of our results is the large upward spike in implied asset volatility in mid-to-late 1988. This may be associated with the tightening of monetary policy (as reflected in increasing short term interest rates, such as the official cash rate) which began about April 1988. Less pronounced increases in asset volatility beginning in late 1983 and early 1984 also coincide with increased short term rates. In comparison with results found for the US (see, for example, Levonian (1991b)), Australian banks do not appear to be noticeably more risky than US banks.

Again, the pattern of asset volatility is greatly influenced by the three largest banks in the sample. The patterns for individual banks differ to some degree. However, in no case is there a significant increase in bank riskiness over the sample period.

While we are concerned with measuring absolute risk rather than systematic risk, the finding that the variance of bank assets has not increased over the 1980s is consistent with the results of Harper and Scheit (1991). Harper and Scheit use the Capital Asset Pricing Model to assess changes in the stock market's perceptions of the riskiness of banks, relative to the total share market, over the 1980s. They find that banks' systematic risk has not increased since deregulation. This is in contrast to the widespread perception that banks have become riskier since deregulation, as evidenced by increases in banks' provisions for bad and doubtful debts as a percentage of assets (see, for instance, Ackland and Harper (1990)).

One reason for our finding might be that deregulation has enabled banks to limit risk by holding more diverse portfolios. For instance, banks' holdings of foreign currency assets and liabilities more than doubled as a proportion of total Australian dollar assets and liabilities between 1984 and 1992. It may also be that risk management techniques have become more sophisticated, or that banks are taking increasing advantage of newly developed hedging instruments.

5.3 The Deposit Guarantee Liability

Annual averages of the sum, across banks, of the liability that would arise in the event that the public sector undertook to guarantee deposits are presented in Table 1. As noted above, calculation of the insurer's contingent liability assumes that all liabilities, not just deposits, are insured. Apart from a couple of large values associated with the share market crash and the sharp escalation of asset volatility in 1988 the public sector liability is negligible, particularly in comparison to results obtained for the US (Levonian (1991b)). This is probably partly a reflection of the fact that banking licence value is considerably higher in Australia than in the US. Movements in asset volatility appear to be offset by increases in bank capital, so that the deposit insurance liability is, in effect, covered by shareholders' capital.

5.4 Varying c and ϕ

Comparing the results for the four cases enables investigation of the sensitivity to changes in the assumed values of c and ϕ. Graphs 1 and 2 show that altering the closure threshold and the licence value makes little difference to the pattern of movements in asset volatility and market capital-asset ratios over time, although the level of the two measures is affected to some degree. Variations in the assumed closure threshold (comparing case 1 with case 2, and case 3 with case 4) had essentially no impact on estimated capital-asset ratios. Increasing the licence value from five to six percent causes a discrete drop in the market capital-asset ratio of about one percent. This is to be expected as the capital ratio is defined exclusive of the licence value. Since a higher licence value should make bank stock more valuable, all else equal, a given market value of equity can only be consistent with a lower market value of assets, and hence, a lower market value of bank capital. Given that varying ϕ, in particular, does affect the results noticeably, a degree of caution is warranted when interpreting the results.

Raising the licence value from five to six percent increased asset volatility slightly. Lowering the closure threshold from −0.02 to −0.05 causes estimated asset volatility to be a little higher.

Table 1 demonstrates that earlier assumed closure (that is, higher closure thresholds) reduces the contingent liability, although this effect varies somewhat according to the licence value. Increasing the licence value also reduces the public sector liability. This result follows from the assumption that a bank's licence value is used to offset, to some degree, the payout to depositors following the bank's failure. In both case 2 and case 4, the closure threshold is such that the bank is only closed when the licence value is completely exhausted by the bank's losses. With this late closure assumption higher licence values reduce risk to the insurer, but only by a very small amount. The sensitivity of the contingent liability to assumptions about c and ϕ make it inappropriate to attach much weight to the specific dollar amounts of the liability. However, the pattern over time is roughly similar across the four cases.

5.5 Correlation Between Risk and Capital

Regulatory guidelines phased in from August 1988 explicitly require banks to raise more capital as they move into riskier areas of activity. Graph 3 shows a scatter plot of the annual averages of asset volatility against the market capital-asset ratio for each bank and each year using the estimates derived under case 3. (Three outlying observations have been omitted for clarity.) It can be seen that, with the exception of 1987 which includes the effect of the share market crash, there is an overall positive relation between the two variables. To investigate the relation between risk and capital we regress the capital-asset ratio on asset volatility using cross section and time series data separately. Since estimation is by way of ordinary least squares, the results are only a measure of simple, contemporaneous correlation. Research in progress by the authors conducts more informative tests of the relationship between risk and capital, pooling the data across banks and across time and using vector autoregression estimation to determine the causal relation between the two variables, rather than simply considering contemporaneous correlation.

Table 2 presents the results of regressing the capital-asset ratio against estimated asset volatility for each year using a cross section of annual averages of individual bank results. Apart from 1987, 1989 and 1991 all estimates of the coefficient on the market capital-asset ratio are significantly positive, implying that at any given point in time banks with riskier assets tend to maintain higher capital ratios. The 1987 share crash was an exception in that increased variation in equity values, and hence measured asset volatility, was associated with a large fall in market value. Measured capital fell as a result of the rapid fall in banks' share prices, at the same time that asset volatility rose, reversing the positive relationship found for other dates.

Table 3 shows time series regressions for each bank. Here the nature of the relation between asset volatility and bank capital is less clear. In four cases out of eleven the relation is significantly positive, while in another four cases the coefficient is found to be strongly negative. However, inspection of the data reveals that the instances of negative and insignificant results are largely the result of the high asset volatility combined with low (or even negative in some cases) market capital levels during late 1987. Thus it does appear that, for each bank, an increase in the riskiness of the bank's assets generally is associated with an increase in the bank's capital.

Footnotes

During the last six months of 1987 the capital-asset ratio fluctuated between −3.4 and 1.1 percent and asset volatility reached a peak of about five percent. [15]

Equity is equal to the number of shares on issue multiplied by the share price, whereas capital (the numerator of the capital-asset ratio) is the difference between the market value of assets and book liabilities. [16]