RDP 9216: The Evolution of Corporate Financial Structure: 1973–1990 4. The Relationship between Financial Structure and the Cyclicality of Output and Earnings

The results in Section 3 indicate that the majority of firms significantly altered their financial structure during the 1980s. In this section, we explore the issue of whether changes in financial structure were concentrated in particular industries. This section also discusses the relationship between the evolution of various ratios and the volatility/cyclicality of industry output and earnings. Detailed examination of the relationship between leverage and firm characteristics is undertaken in Lowe, Morling and Shuetrim (1993).

Given expected returns, the risk that a firm is unable to meet its obligations out of current earnings increases with the extent of its leverage and with the volatility of its cash flows. As a result, extremely risk averse firms are likely to have lower leverage, as are firms with highly volatile or cyclically sensitive earnings. If shocks to individual firms are idiosyncratic (that is, if the shocks are uncorrelated across firms) then a widespread increase in leverage, similar to that which took place in the 1980s, is unlikely to have important implications for the growth path of the macro-economy. However, in many cases, major shocks are highly correlated across firms. For example, unexpectedly tight monetary policy is likely to cause a slowdown in economic activity and this slowdown is likely to reduce profits for a wide cross-section of firms. If, as a result of the slowdown, firms wish to increase their interest cover and reduce their leverage, a period of low aggregate investment might ensue as firms become less able to undertake additional debt financed investment. This would imply that firms are foregoing investment projects with positive net present values. The problem is potentially more severe if the increase in leverage is predominantly concentrated in firms with volatile or cyclically sensitive earnings. On the other hand, firms whose profits are relatively insensitive to the business cycle are less likely to require balance sheet reconstruction following an adverse macro-economic shock.

There is a growing literature on the links between the financial structure of firms and the evolution of the macro-economy. This literature is discussed in some detail in Lowe and Rohling (1993). They identify three key transmission mechanisms through which the state of corporate balance sheets can affect the evolution of the business cycle. The first is that emphasised by Bernanke and Gertler (1990). As a firm's leverage increases and the probability of insolvency rises, financial intermediaries become less willing to lend to the firm, even if it proposes a project with a positive net present value. Second, if corporate insolvencies get translated into the balance sheets of financial institutions, the equity of those financial institutions is reduced. The reduction in equity may result in the financial institutions becoming less able to undertake risky lending. Some form of lending institution induced credit squeeze may result. Third, and perhaps most importantly, the incentives of managers are a function of the financial structure of the firm. The higher probability of financial failure may reduce the incentives for risk-averse management, with firm-specific capital, to undertake further ex-ante profitable, but risky, investment.

Recent research has examined the link between changes in leverage and earnings volatility. Seth ((1990) and (1991)) calculated correlations between the degree of cyclicality in industry earnings and both the extent of industry leverage, and its rate of change, for a range of countries including Australia. His results suggest that, in the United States, increases in leverage have been concentrated in those industries with cyclically sensitive earnings. No such result was found for Australia; the average annual correlation between cyclicality and leverage being less than 0.1 over the sample period. Seth found a similarly small average annual correlation between cyclicality and the growth rate of leverage in Australia. Seth also analysed the relationship between earnings cyclicality and both the level and growth rate of interest cover. He concluded, again on the basis of annual average correlation coefficients, that there was no evidence that cyclically sensitive sectors in Australia had either lower levels of interest cover or more rapid declines in interest cover.

The conclusions reached by Seth require some qualification. His results are based on correlations between the average leverage of firms in each of the 10 industry groupings and the degree of cyclicality of each industry's earnings. Separate correlations coefficients between industry leverage and industry cyclicality are calculated for each of 10 years and the correlations are then averaged. For the Australian firms, the correlation between leverage and cyclicality vary considerably over time; in 1983, the correlation was −0.81 while in 1987 it was 0.39. While the correlations are generally positive, indicating that industries with cyclical earnings had higher leverage, the average correlation is essentially zero. To a large extent this reflects the high negative correlation in 1983.

Lee (1990) also considered the relationship between leverage and cyclicality but concentrated his study on the United States. Again, Lee classified industries into cyclical and non-cyclical groups based upon the correlation, within each industry, between firm cash flow and the business cycle. Median debt-asset ratios indicated that the leverage of firms in cyclical industries was marginally lower than that of firms in the more stable industries. Further, he found that the pattern was similar in 1978 and in 1987 and that the relative debt-asset ratio rankings of industries within their cyclicality groupings were fairly stable over the time period studied.

To explore the Australian experience more deeply, we compare various measures of industry cyclicality and industry volatility to weighted average and median industry financial ratios. We begin by classifying the firms in our database into industry groups. The choice of industries is restricted to the intersection of industry classifications in our database and those in the National Accounts data. These are the manufacturing, mining, wholesale trade, retail trade and service sectors. Most firms are classified into one of these five industry groupings. Given that many firms are highly diversified, the classifications are far from perfect. Of the 110 firms, five were excluded from the analysis as we felt that they could not sensibly be put into any category. The classifications are given in Appendix 1.

Having classified each firm into one of the five industry groups, we then examine the degree of volatility and cyclicality of industry earnings and output. Industry output is derived from the National Accounts data and average earnings is based upon each individual firm's financial statements. Finally, we examine the relationship between changes in leverage, interest cover and the ratio of trade creditors to debt for each of the industry groups and the volatility/cyclicality of the industries' earnings/output.

Various methods are used to construct measures of cyclicality and volatility. The simplest measure of volatility is the standard deviation of the growth rate of industry output. This is calculated using the percentage change in quarterly industry GDP measured at 1984/85 prices. These standard deviations are shown in the first column of Table 1. Table 1 also contains the coefficient of variation (the standard deviation divided by the mean). Higher volatility of industry output does not necessarily imply a greater riskiness if the average growth rates of firms in the industry are also higher. The coefficient of variation attempts to control for the different rates of growth experienced in various industries between 1973 and 1990.

Both the standard deviation and the coefficient of variation are measures of the volatility of industry output. It is also interesting to examine the extent to which industry output and profitability vary with the business cycle. The first method used to obtain a measure of industry cyclicality involves regressing the quarterly percentage change in real industry GDP on a constant and the quarterly percentage change in total GDP^[11].

If changes in total GDP are uncorrelated with changes in industry GDP, the coefficient on total GDP (β) will be zero. Typically, those industries with coefficients greater than one are said to be cyclically sensitive while those with coefficients less than one are said not to be cyclically sensitive. This distinction is somewhat arbitrary in that it is made relative to the normal level of cyclicality defined by the percentage changes in total GDP.

A second measure of cyclicality can be obtained by calculating the correlation coefficients (ρ) between changes in industry output and changes in GDP. These correlation coefficients equal β multiplied by the ratio of the standard deviation of the total GDP growth rate over the standard deviation of the industry output growth rate. Both the estimates of β and ρ are presented in Table 2 along with their standard errors.

A similar method is used to gauge earnings cyclicality. This measure is estimated by regressing the industry rate of return on a constant and the gap between actual output and potential output.

The industry rate of return is based upon an individual firm's financial statements^[12]. Potential output is estimated using a Hodrick-Prescott (1980) filter which generates a non-linear trend from the actual GDP series^[13]. The Hodrick-Prescott filter minimises a function of the sum of squared deviations of the trend value from the logarithm of actual GDP and a penalised sum of the squared changes in trend in each quarter. Again, the size of the coefficient on the GAP variable is an indication of the cyclicality of an industry's average rate of return. Industries whose rates of return on assets vary substantially with the business cycle should generate higher estimates of γ than should industries whose profitability is largely independent of the business cycle. The estimated coefficients are reported in the third column of Table 2 along with their standard errors^[14].

The results in Table 1 show that the standard deviation of the growth rate of the output of the mining industry is considerably higher than the standard deviation for the other sectors. On the basis of the standard deviation measure of volatility, wholesale and manufacturing are the next two most volatile sectors, followed by the retail sector and finally, the service sector. Using the coefficient of variation measure, manufacturing has the highest measure of volatility with a coefficient of 10.6 followed by wholesale trade with a coefficient of 7.2 and mining with 4.2. Retail trade and service are considerably more stable with coefficients of 3.0 and 1.6 respectively. It should be noted that the relative ranking of the mining industry alters between measures. During the period from 1973 to 1990, the output of the mining industry grew much more rapidly than that of the manufacturing and wholesale sectors. As a result, the coefficient of variation of the growth rate of the mining sector is less than that of the manufacturing and wholesale sectors.

Given our definition of a cyclical industry, the measure of cyclicality obtained from estimating equation (5) shows that the mining, manufacturing and wholesale sectors are cyclical while the retail and service sectors are relatively insensitive to the business cycle. The retail industry appears to exhibit the least degree of cyclicality, followed by the service industry, both of which have coefficients significantly less than one. Manufacturing and wholesale trade, on the other hand, are clearly cyclical with coefficients significantly greater than one. The coefficient on mining is also greater than one but not significantly so^[15]. The correlation coefficients show broadly similar rankings except that the mining sector appears much less cyclical than the results in column one of Table 2 suggest. This reflects the relatively large standard deviation of the growth rate of the mining sector.

The regressions of industry return on the output gap yielded relatively imprecise estimates of the degree of cyclicality. Of the five parameters estimated, only the coefficients in the equation for the manufacturing and retail sectors are significantly different from zero at the 5 percent level of significance. The service industry coefficient is actually negative unlike those of the other industries. However, the insignificance of the service industry coefficient prevents any special interpretation of the result.

The above results suggest the following conclusions. Both the retail and the service sectors have output that is relatively stable and both sectors are relatively insulated from the business cycle. In contrast, output of the manufacturing and wholesale sectors is relatively volatile and influenced more heavily by the business cycle. The mining sector has the most volatile output but the volatility in its output is less directly associated with the business cycle than is the case for the manufacturing or wholesale industries.

The ranking of the five industry groups can be compared to the industry debt-asset, interest cover and creditors to debt ratios to determine whether the firms in the more cyclical and the more volatile industries have lower leverage and have exhibited more restraint in their debt expansion. The dividend pay-out ratio was also considered but no substantial industry specific effects were detected. To allow an examination of the relationship between cyclicality/volatility and the levels and changes in the various ratios over time, the weighted averages and medians, by ratio, for each of the five industries are presented in Tables 3 through 8.

Of the relatively volatile industries, manufacturing is clearly the one that has made the most significant changes to its financial structure over the 1980s. The results in Table 3 show that the weighted average debt-asset ratio for manufacturing increased from 0.46 in 1973 to 0.61 in 1990. The median debt-asset ratio for manufacturing in Table 4 shows a similar rise.

Associated with the increase in leverage of manufacturing was a pronounced decline in interest cover. The weighted average interest cover (Table 5) fell from 8.9 in 1973 to 3.6 in 1990 while the median interest cover (Table 6) fell from 9.9 to 4.0 over the same period. Although most of the financial structure adjustment may have reflected a movement from a constrained position to an unconstrained position, the severity of the adjustment in the manufacturing sector suggests that some firms may have increased their leverage excessively.

The weighted average debt-asset ratio for mining actually fell between 1973 and 1990 after a brief increase in the mid 1980s. However, this fall is mainly associated with a single firm that had total assets greater than the sum of total assets across all other mining firms. Given the dominance of the single firm, the median debt-asset ratio may provide a more accurate picture of the industry as a whole. The median figures show considerable volatility, increasing significantly between 1973 and 1975 before falling in the early 1980s and then increasing again through the remainder of the decade. The low figure for 1980 (both the median and the weighted average) reflects the equity raisings associated with the increase in investment in the mining sector in the early 1980s.

As is the case for the debt-asset ratio, the interest cover ratio for the mining sector is volatile. Although it appears that interest cover in the mining sector has not experienced the same sustained decline as in other sectors, the mining boom in 1980 raised the median interest cover to 12.4. The 1982/83 recession, however, caused it to fall to a low of 5.36. By 1990 interest cover had recovered to be similar to the cover experienced during the first half of the 1970s.

The results in Tables 7 and 8 show that, of the five sectors examined, the mining sector has made the least use of trade credit. In 1990, such credit accounted for less than 10 percent of total debt in the mining industry. However, like the other industry groups, the reliance on trade credit did increase over the 1970s and then fell over the 1980s.

The results for the wholesale sector need to be interpreted with some caution. Although the wholesale firms faced the highest leverage and lowest interest cover during most of the sample period, it is difficult to compare these results with those of other industries because three of the six wholesale firms in our sample are subsidiaries of Japanese multinationals. The consolidated risk position of the parent company and its subsidiaries should be considered rather than the individual accounts of the subsidiary. Thus, the high leverage and low interest cover of the wholesale industry may not be representative of the subsidiaries' true financial security.

Traditionally, the retail sector has been more highly geared than the manufacturing sector. However, the increase in leverage in the manufacturing sector during the 1980s has meant that, more recently, the two industry groupings have been similarly geared. Over the sample period, the weighted average debt-asset ratio of firms in the retail sector increased from 0.50 to 0.60 while the weighted average interest cover ratio fell from 10.9 to 3.5. The results in Tables 7 and 8 also show that the retail firms have typically been heavy users of trade credit. In 1980, creditors accounted for 46 percent of the total debt of the retail firms. This share had fallen to 29 percent by 1990. The results also show that between 1980 and 1990, there was little change in the gearing of the retail industry. This may have reflected the changing composition of total debt away from trade creditors towards other forms of debt.

Table 3 shows that the weighted average debt-asset ratio for the service industry increased marginally from 0.64 to 0.67 over the period. A possible reason for the limited expansion of debt by the service sector is the fact that its leverage was already high in comparison to the other industries. Accompanying the relatively minor debt expansion that occurred between 1973 and 1990 were significant declines in the weighted average and median interest cover ratios from 6.6 to 2.9 and from 9.6 to 3.6 respectively. The declines primarily reflect a substantial increase in the average interest rate paid. There was also a slight decline in the ratio of earnings to total assets.

Of the five industry groups considered in this section, developments in the manufacturing sector are of potential concern. The evidence suggests that the increases in leverage and the decline in interest cover were very pronounced for the manufacturing sector. The manufacturing sector also appears to have the most cyclically-sensitive output^[16]. A significant increase in leverage within a cyclically-sensitive sector is exactly the situation in which increasing average leverage is likely to affect the evolution of the macro-economy. An adverse shock to demand places the highly geared firms under considerable financial distress. This distress may well lead to a period of relatively low investment as companies attempt to retain earnings in an effort to improve their financial structure. The likely consequence is a slower recovery from an adverse macro-economic shock. This is not to say that high debt is the only factor that can slow investment recovery. Most notably, excess capacity and slack demand, by reducing the number of investment projects with positive net present value, will also reduce the level of corporate investment.

Footnotes

Both the industry aggregate and total aggregate measures of GDP are seasonally adjusted series measured at constant prices from the Australian National Accounts, catalogues 5206.0 and 5222.0. [11]

A weighted average measure of firms' rates of return was used for each industry. This was calculated as the sum of net profits across the firms in an industry divided by the sum of total assets across the same group of firms. [12]

The Hodrick-Prescott procedure was run with the non-linearity penalty parameter (λ) set at 1,600, the value favoured by Kydland and Prescott (1990). The original data series is quarterly real GDP with 1987 as the base year and runs from March 1964 to December 1991. To generate annual observations, a weighted average of the observed quarterly gaps over eight quarters is used. The eight quarters capture all possible observations relating to the financial statements in one year and the weights reflect the fraction of financial statements reported in each quarter. [13]

Note that the standard errors of the estimated γ coefficients are not directly comparable between industry equations in this table. This reflects the fact that, when averaging the data over an increasing number of firms, that part of the variance caused by variability in firm specific factors becomes increasingly small. Those industries with a large number of firms would, thus, be expected to exhibit a smaller total error variance. [14]

All significance tests are done at the five percent level using a one tailed test. [15]

The internationalisation of the Australian manufacturing sector may well have made the industry, as a whole, less sensitive to the Australian business cycle in recent years. In this case, the measured cyclicality of the manufacturing sector may overstate the actual degree of cyclicality in recent years. [16]