Capital structure dynamics with interrelated adjustment: Australian evidence.

Keywords: CAPITAL; DYNAMICS; ADJUSTMENT; AUSTRALIAN.

1. Introduction

While financial economists have devoted considerable attention to empirically testing alternative theories of optimal capital structure, relatively little research has focused on explaining the dynamics of a firm's capital structure in the presence of transactions costs. Indeed in most empirical studies it is assumed that the firm's capital structure, when averaged over a three to five year period, is its desired or optimal capital structure. This averaging procedure is not only ad hoc and subject to measurement error, but it has the further disadvantage of precluding any study of the capital structure adjustment process over time. This omission is surprising in light of Myers' (1984, p. 587) conjecture that `large adjustment costs could possibly explain the wide variation in debt ratios, since firms would be forced into long excursions away from their initial debt ratios'.

The limited number of empirical studies which have considered capital structure dynamics may be classified into two groups depending on whether they utilise cross-sectional or time-series data. Fischer, Heinkel and Zechner (1989) use cross-sectional data in testing their model of the optimal dynamic recapitalisation policy of a firm with exogenous investment decisions and in the presence of transactions costs. Generally their evidence is consistent with the hypothesis that a firm's debt ratio range, defined as the difference between its maximum and minimum debt ratio over a 34 quarter sample period, is related to adjustment or transactions costs in recapitalisation. However a shortcoming of this methodology is that by focusing on the range between the extreme values of a firm's capital structure it neglects short-run variability in capital ratios.

The second group of studies of capital structure dynamics utilise pooled time-series/cross-sectional data (see, eg, Taggart 1977; Marcus 1983; Jalilvand & Harris 1984; Sharpe 1991a, 1991b; Sharpe & Pooley 1991). In the presence of adjustment costs firms are assumed to gradually adjust their capital (debt) ratio at a constant rate so as to eliminate deviations between their optimal (or desired) and actual capital (debt) ratio. This simple Koyck partial adjustment model is augmented by permitting exogenous transitory factors, referred to as innovations, to impinge on capital structure decisions.

The augmented Koyck model, however, has several shortcomings. Thus Marcus (1983, p. 1,220) acknowledges that the constant speed of adjustment assumption is too rigid and that a more general or flexible adjustment model would be preferable. Jalilvand and Harris, and Sharpe and Pooley attempt to overcome this problem by allowing the partial adjustment parameter to have both a fixed and variable component, the latter reflecting different perceived adjustment costs or benefits of adjusting to capital structure targets. For estimation purposes Jalilvand and Harris simplify their model by assuming that the desired capital (debt) ratio is exogenous, thereby ignoring the literature on the determinants of optimal capital structure while focusing on the variable adjustment mechanism. As the factors assumed to influence adjustment speeds frequently appear as proxies for determinants of optimal capital structures in cross-sectional studies, the Jalilvand and Harris methodology may produce biased estimates. Moreover, attempts to estimate a model incorporating both variable adjustment and the determinants of desired capital (debt) ratios have been unsuccessful because of a severe multicollinearity problem.(1)

A further problem with the Koyck model emerges when it is used to estimate the individual components of a portfolio of liabilities (assets). Because of the total liability (asset) constraint whereby the individual liability (asset) components must sum to the whole, one of the individual liabilities (assets) is chosen as the residual liability (asset) and excluded from the estimation of the system of equations. But when the individual components are estimated using the Koyck model, the regression results are sensitive to the choice of which liability (asset) is used as the residual liability (asset).

Finally we note that, holding total liabilities (assets) constant, the Koyck model assumes that the adjustment of any one liability (asset) is offset by a corresponding and offsetting adjustment in the residual liability (asset). Thus the Koyck model does not allow for interrelated adjustment among the non-residual liabilities (assets). In reality, the adjustment of any one liability (asset) may depend not only on its own gap between desired and actual levels but also on the deviations of other non-residual liabilities (assets) from their desired levels (see Brainard & Tobin 1968; Nadiri & Rosen 1969). This formulation of the adjustment process, which we refer to as the general or multi-variate stock adjustment model, permits a more flexible adjustment process to be estimated than is possible with the simple Koyck model (see Sharpe 1973, 1974). Moreover, the general stock adjustment model has the further advantage of the econometric estimates of the model being invariant to the choice of liability (asset) excluded from the estimation of the system of equations.(2)

Hence the primary objective of this paper is to formulate and estimate a model to explain how firms adjust to their desired capital structure. Specifically we are concerned with: (i) how innovations in investment and retained earnings impinge on the firm's capital structure; (ii) whether the adjustment process is characterised by interrelated adjustment and hence best modelled using the general stock adjustment model rather than the simple Koyck model; and (iii) the speed of adjustment involved in restoring a long-run equilibrium position after an innovation disturbs an existing equilibrium. The latter could shed light on the appropriateness of the common practice in cross-sectional studies of optimal capital structure of averaging the data over a 3-5 year period so as to eliminate disequilibrium effects.

A secondary objective of the study is to test the primary predictions of the theory of optimal capital structure. Prior to the mid 1980s this literature focused on the choice between debt and equity. However, theoretical work by Brick and Ravid (1985), Diamond (1991) and others has stimulated empirical research on the structure of corporate debt including the mix of short-term and long-term debt, the priority structure of debt, and the use of call and sinking fund features.

There has been relatively little interest in these issues by Australian researchers, apart from Chiarella et al. (1992). They apply the static linear structural modelling approach of Titman and Wessels (1988) to explain average debt to equity ratios of 226 Australian firms over the 1980-1982 period and find results broadly similar to those in the corresponding USA study. Our study departs from Chiarella et al. in providing a dynamic and interrelated analysis of capital structure choice of Australian firms over the 1967-1985 period.

The data for this study is drawn from the Australian Graduate School of Management (AGSM) Annual Report File. Given the limitations of the data classifications within the data base, the three components of a firm's capital structure are defined as follows: (i) equity includes share capital, reserves, retained profits and minority interest; (ii) long-term debt is total long-term (deferred) liabilities; and (iii) short-term debt is total current liabilities (see the appendix for definitions). Figure 1 depicts the time series experience of the annual mean of the ratio of each of the capital structure components to total debt and equity for the 164 companies included in the sample.(3)

[Figure 1 ILLUSTRATION OMITTED]

Several features of the data are noteworthy. Firstly, there was a statistically significant fall in the equity ratio, or increase in leverage, over the 1967-1985 period although much of the fall occurred in the first ten years.(4) Secondly, Australian firms rely heavily on short-term rather than long-term debt.(5) Thirdly, the averaging procedure has concealed considerable cross-sectional and time-series variability in the capital structure components of individual firms. Cross-sectional variability is reflected in relatively high standard errors of the yearly mean ratios shown in figure 1. For example, in 1985 the mean and standard errors (in parentheses) of the capital ratios were 0.579 (0.224) for equity, 0.123 (0.129) for long-term debt, and 0.298 (0.178) for short-term debt across all firms. Moreover, the average time series variability of the capital ratios for individual firms over the nineteen year sample is reflected in the standard errors of the individual firm global mean capital ratios of 0.080 for equity, 0.053 for long-term debt, and 0.073 for short-term debt.

2. Model Specification

2.1 A Dynamic Simultaneous Equation Model With Interrelated Adjustment

Noting that for a typical firm i, at the end of period t, total liabilities, [TL.sub.it], is the sum of equity, [E.sub.it], long-term debt, [LTD.sub.it], and short-term debt, [STD.sub.it], then:

(1) [YE.sub.it] + [YL.sub.it] + [YS.sub.it] = 1,

where YE is the equity to total liability ratio, YL is the long-term debt to total liability ratio, and S is the short-term debt to total liability ratio. Our objective is to develop a system of equations comprising behavioural equations for the equity and long-term debt ratios while the short-term debt ratio is determined residually from the identity given by equation 1.(6)

The behavioural equations postulate that changes in the equity (long-term debt) ratio derive from three sources: (i) the firm gradually adjusts its beginning of period equity (long-term debt) ratio to its desired or long-run equilibrium ratio so as to minimise the total of transactions costs and the opportunity cost of diverging from the long-run equilibrium ratio; (ii) because of liability interdependence and differing adjustment costs across liability categories, changes in long-term and short-term debt may impact on the equity ratio (and similarly changes in the equity and short-term debt ratios may impact on the long-term debt ratio)V; and (iii) various transitory factors, or innovations, may cause the equity (long-term debt) ratio to diverge from its desired or long-run equilibrium level. Incorporating these three elements and assuming two innovations, denoted I1 and I2 respectively, the behavioural equations for equity and long-term debt are given by:

(2) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

(3) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

where an (*) indicates the desired or long-run equilibrium level of the variable, and [sub.E][[Lambda].sub.K] and [sub.L][[Lambda].sub.k] for k = 0, 1, 2, 3, 4 are estimated parameters. While the own-adjustment parameters, [sub.E][[Lambda].sub.0] and [sub.L][[Lambda].sub.0] are expected to be positive, the signs of the cross-adjustment parameters, [sub.E][[Lambda].sub.1], [sub.E][[Lambda].sub.2], [sub.L][[Lambda].sub.1] and [sub.L][[Lambda].sub.2] are uncertain. If the cross-adjustment parameters are insignificantly different from zero then the model in equations 2 and 3 collapses to an innovation augmented Koyck partial adjustment model.

Because of identity 1 any factor influencing the desired or actual equity ratio must have an equal and offsetting influence across the two debt ratios. Thus, in general terms, factors influencing the desired equity ratio should also appear as explanatory variables in the desired debt ratios, and vice versa. We therefore assume that the desired or optimal equity and long-term debt ratios of a firm depend linearly on a common set of J variables, denoted [X.sub.j] for j = 1 to J:

(4) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

(5) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

where the [[Beta].sub.0] and [[Beta].sub.j] are estimated parameters. Then, substituting equations 4 and 5 into equations 2 and 3 and noting that [Delta]YS = - ([Delta]YE + [Delta]YL) we have:

(6) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

(7) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

A difficulty with the multi-variate stock adjustment model given by equations 6 and 7 is that the econometric estimates of the model produce short-run parameter estimates which are non-linear functions of the adjustment parameters and of the parameters of the long-run equilibrium or desired equations. The difficulty lies in calculating standard errors and thus significance tests for the derived long-run parameters, the [[Beta].sub.j], as they typically do not possess finite sample moments.

A simple transformation, as suggested by Bewley and Fiebig (1990), overcomes this difficulty. Re-arranging equations 6 and 7 we obtain:(8)

(8) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

(9) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

Thus regression estimates of equations 8 and 9 provide direct estimates of the long-run equilibrium determinants of capital structure, the [[Beta].sub.j], and their standard errors.(9)

An important property of equations 8 and 9 arises from their having an identical set of explanatory variables. That is, econometric systems estimators, such as seemingly unrelated regression, collapse to ordinary least squares estimates of the individual equations. Moreover, the estimates of the model will be invariant to the choice of liability to act as the residual liability in the system of equations.

2.2 The Innovation Vector

In modelling firms' adjustment of their capital structure, Jalilvand and Harris (1984) assume that investment and retained earnings are innovations and that the adjustment process is identical, though opposite in sign, for each of the innovations, This approach is derived from the sources and uses of funds identity. On the other hand Marcus (1983) identifies three types of innovations in his market value based study of USA bank capital decisions: unanticipated income, capital gains and deposits. Moreover, he allows each type of innovation to impact differently on capital structure decisions.

The primary use of funds in our model is investment, proxied by the change in total assets of a firm. On the other hand, the sources of funds include long- and short-term debt, retained earnings and new issues of equity. Because of our focus on the financing decision, we assume that investment and retained earnings are exogenous.(10) Thus, in the short-run, innovations in a firm's investment and retained earnings impinge directly on its financing decision through the sources and uses of funds identity.(11) However, we follow Marcus in: (i) focusing on unanticipated components of these variables as proxies for the innovations; and (ii) allowing different adjustment processes for each of the innovations.

Scaling each of the innovations by beginning of period total assets, [TA.sub.t-1], which equal total liabilities, [TL.sub.t-1], the innovation variables used in the estimation of the model become the unexpected growth in total assets (liabilities) and the unexpected retained profits to lagged total assets (liabilities) ratio. We then have:

(10) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

(11) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

where `u' denotes unexpected, and NIATD is net income after taxes and dividends. Unanticipated new investment could be initially financed by short- and long-term debt with their lower transaction costs vis-a-vis equity. Consequently, a positive innovation in investment is likely to reduce the equity ratio and increase the long- and short-term debt ratios. Thus in equations 8 and 9 we anticipate:

(12) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

Furthermore, as retained earnings feed directly into shareholders funds and our definition of equity includes retained profits as well as share capital and reserves, then positive innovations in retained profits will initially increase the equity ratio with offsetting reductions in debt ratios so that:

(13) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

2.3 Desired Equity and Debt Ratios

Although the theoretical literatures explaining optimal corporate leverage and optimal maturity of corporate debt have developed in somewhat independent strands, the underlying determinants of optimal leverage and debt maturity are nevertheless quite similar. In this section we briefly examine the principal hypotheses relating to these determinants and, following Barclay and Smith (1993, 1995), group those hypotheses into three non-mutually exclusive sets: tax hypotheses, information asymmetry, and contracting cost hypotheses. It should be noted that a prediction from the debt maturity literature that debt maturity is positively related to a variable [X.sub.j] requires the coefficient of [X.sub.j] in the long-term debt equation to be greater than its coefficient in the short-term debt equation. That is, [sub.L][[Beta].sub.j] - [sub.S][[Beta].sub.j] [is greater than] 0 or equivalently [sub.E][[Beta].sub.j] + 2* [sub.L][[Beta].sub.j] [is greater than] 0.

2.3.1 Tax Hypotheses Our model incorporates proxies for the firm's expected interest tax shield, X1, and expected non-debt tax shield, X2. Miller (1977, p. 267) showed that the tax advantage of debt vis-a-vis equity depends on the marginal personal tax rate on interest, the tax rate on capital gains, and the effective corporate income tax rate. We focus on the firm's expected effective tax rate with the theory anticipating an inverse relationship to the equity ratio. Moreover, drawing on Brick and Ravid's work, Barclay and Smith (1993), and Guedes and Opler (1994) argue the interest tax shield proxy will be directly related to debt maturity when the yield curve is upward sloping (as it was in Australia through most of our sample period). Hence the theory predicts:

(14) [sub.E][[Beta].sub.1] [is less than] 0 and [sub.L][[Beta].sub.1] - [sub.S][[Beta].sub.1] [is greater than] 0.

In De Angelo and Masulis' (1980) framework non-debt tax shields, such as depreciation allowances and investment tax credits, are substitutes for interest expenses because of their capacity to reduce the firm's effective marginal tax rate on interest deductions. Hence:

(15) [sub.E][[Beta].sub.2] [is greater than] 0 and [sub.L][[Beta].sub.2] - [sub.S][[Beta].sub.2] [is less than] 0.

2.3.2 Information Asymmetry Given information asymmetries, management may signal the firm's quality to the market through capital structure decisions. A high quality, though undervalued firm, may signal its high expected future profitability by increasing its leverage (see Blazenko 1987) and thereby avoid the need to issue new equity at undervalued prices. On the other hand, Myers and Majluf (1984) show that, in the presence of asymmetric information, firms may choose not to issue underpriced new securities and will thereby forego positive NPV projects. As firms with greater access to retained profits may avoid these costs by having greater financial slack, a positive relationship is predicted between profit retention and the equity ratios. We follow Mitchell (1991) in proxying asymmetric information by the firm's expected profit retention ratio, X3, with high quality undervalued firms being associated with high expected retained earnings. Given the offsetting influences described above, the sign of [sub.E][[Beta].sub.3] is uncertain.

The signalling models also have implications for debt maturity (see Flannery 1986; Kale & Noe 1990; Barclay & Smith 1993). Because the value of long-term debt is more sensitive than short-term debt to changes in firm value, a high quality undervalued firm will prefer to issue the less-underpriced short-term debt so that:

(16) [sub.L][[Beta].sub.3] - [sub.S][[Beta].sub.3] [is less than] 0.

2.3.3 Contracting-Cost Hypotheses Myers (1977) shows how, in the presence of fixed rate debt, conflicts may arise between debt holders and stockholders over the firm's investment decisions. The conflict, which is more likely in firms with growth options, may lead to an under-investment problem if the benefits of new investment accrue primarily to debt holders. However the under-investment problem is reduced by the firm using more equity and less debt, by imposing restrictive covenants on the debt, or by shortening the maturity of the debt. Alternatively, Titman and Wessels (p. 4) argue that where firms issue short-term rather than long-term debt, as is common in Australia because of poorly developed long-term debt markets, the agency cost is mitigated so that equity ratios could be inversely related to growth opportunities. Furthermore, an inverse relationship is also suggested by the dynamics of the pecking order theory which implies that `an unusually profitable firm in an industry with relatively slow growth (few investment opportunities) will end up with an unusually low debt-to-equity ratio' (Copeland & Weston 1988, p. 507). We use the firm's expected growth in total assets, X4, as a proxy for growth opportunities with the sign of [sub.E][[Beta].sub.4] uncertain while:

(17) [sub.L][[Beta].sub.4] - [sub.S][[Beta].sub.4] [is less than] 0.

Where a firm's assets are collateralisable, shareholders may be restricted in their ability to appropriate wealth from debt holders suggesting a negative relationship between collateralisable assets, X5, and equity ratios. Generally collateralisable assets have long maturities and Myers (1977) suggests firms match debt with asset maturities to lower agency costs of debt while Diamond argues that duration matching reduces liquidity risk. Hence we expect:

(18) [sub.E][[Beta].sub.5] [is less than] and [sub.L][[Beta].sub.5] - [sub.S][[Beta].sub.5] [is greater than] 0.

If bankruptcy or financial stress reduces shareholder wealth, then higher risk firms would avoid debt so as to reduce interest payments and the probability of default. On the other hand, lenders are concerned with the moral hazard problem associated with the issue of risky debt as the high risk premium may induce the borrower to undertake excessively risky projects. Indeed Diamond argues this moral hazard problem results in risky firms being screened out of the long-term debt market. Furthermore, Barclay and Smith (1993) note that as banks and other lenders have a comparative advantage in monitoring the activities of risky firms and bank debt has relatively short maturity then more risky firms will use relatively more short-term bank debt than less risky firms.

We use three proxies for firm risk: (i) firm size, X6; (ii) liquidity, X7; and (iii) volatility of earnings, X8. If large firms are of lower risk than small firms because of greater diversification, then firm size will be negatively related to equity ratios and positively related to debt maturity. However firm size may also proxy transactions costs. Thus large firms may be able to raise new equity capital and long-term public debt with less difficulty and lower transactions costs than small firms, so that firm size would be positively related to both equity ratios and debt maturity. Thus the relationship between firm size and equity ratios has an uncertain sign while:

(9) [sub.L][[Beta].sub.6] - [sub.S][[Beta].sub.6] [is greater than] 0.

Our proxy for liquidity risk, the current ratio, is inversely related to liquidity risk so that the current ratio will be negatively related to equity ratios and positively related to debt maturity. However the current ratio is likely to be inversely correlated with collateralisable assets and, consistent with the discussion above, this leads to an expectation of a positive relationship with equity ratios and a negative relationship with debt maturity. With these confounding influences the signs of both [sub.E][[Beta].sub.7] and ([sub.L][[Beta].sub.7] - [sub.S][[Beta].sub.7]) are thus uncertain.

Finally, the volatility of earnings, X8, is expected to be positively related to equity ratios and inversely related to debt maturity:

(20) [sub.E][[Beta].sub.8] [is greater than] and [sub.L][[Beta].sub.8] - [sub.S][[Beta].sub.8] [is less than] 0.

3. Data and Estimation Technique

The data used in this study was collected from the AGSM Annual Report File but, because of incomplete data in the early years, the study was limited to those companies with complete data encompassing the period 1966-1985. Only 165 companies met this criteria and one of those companies was subsequently excluded from the sample because it had an indeterminate liquidity ratio for 1984. With the pooling of the annual and cross-sectional observations, the regression analysis utilises 3,116 observations, there being nineteen useable annual observations for each of the 164 companies.(12)

The appendix contains definitions of the variables used in the regression analysis. Being interested in managerial initiated changes in capital structure, the equity and debt ratios are defined in book values. As noted by a referee, while book values are inconsistent with the capital structure theory, it avoids having to separately model share price induced and managerial initiated changes to capital structure. Moreover, as long-run variables involve expected values and innovations unexpected values, it is necessary to assume an expectations generating mechanism. We adopt a simple model in which management uses a no change forecast in forming their expectations.(13) Denoting expected values by `e' then for variable Xj we have:

(21) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

The surprise or unanticipated component, denoted by `u' is then given by:

(22) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

Judge et al. (1985, ch. 11) discuss alternative estimation methods for pooled time series and cross-sectional data including the fixed effects model and the error components model.(14) The fixed effects model is preferred for the purposes of this study for two reasons. Firstly, it is less likely to produce biased estimates when unmeasurable individual firm attributes are correlated with the measurable timevarying explanatory variables in the model. Secondly, in a simultaneous equation system the error components model is difficult to implement and is sensitive to the choice of equation used to decompose the residuals.

Although the model involves all interrelated system of equations, that is, the behavioural equations 8 and 9 and identity 1, a systems estimator is not required because the behavioural equations contain an identical set of explanatory variables. However, Bewley and Fiebig show that in order for the parameter estimates of the long-run coefficients and adjustment coefficients to mirror those in equations 6 and 7, it is essential to use an instrumental variable estimator on equations 8 and 9. Moreover, the set of instruments must include each of the (individual firm mean differenced) exogenous variables in the equations (ie, the constant, XJ for j = 1 to 8, and the two innovations) and the two lagged dependent variables, [YE.sub.i,t-1] and [YL.sub.i,t-1], which serve as instruments for the changes in the dependent variables.15

Finally, preliminary regression estimates suggested the presence of heteroscedastic errors when the Breusch and Pagan Lagrange Multiplier test was applied. Hence, the final model was estimated using weighted regression with the residual variances assumed to be proportional to the inverse of firm size, X6.

4. The Results

Table 1 reports the fixed effects weighted instrumental variable estimates of the model. Although the model was initially estimated with the short-term debt equation omitted, we have included the result for short-term debt which was obtained in a subsequent estimation of the model with the long-term debt equation excluded.

Table 1 Fixed Effects Weighted Instrumental Variable Estimates of the Innovation Augmented Interrelated Adjustment Model -- Capital Structure of Australian Firms 1967-1985

Independent Variables                           Equity YE
                                         Estimate    t-statistic(1)

Innovations
Unexpected Investment (I1)                -0.2611        -14.25(*)
Unexpected Retained Profit (/2)           0.8476           6.97(*)

Adjustment Mechanism
Change in Equity ([Delta]YE)              -3.8313        -12.93(*)
Change in Long-Term Debt ([Delta]YL)      0.3866           1.76
Long-Run Variables
Interest Tax Shield (X1)                  -0.0297         -0.92
Non-Debt Tax Shield (X2)                  -0.0001         -0.46
Profitability (X3)                        1.0678           7.10(*)
Growth Opportunities (X4)                 -0.3460        -13.34(*)
Collateral Value of Assets (X5)           -0.0254         -0.65
Firm Size (X6)                            -0.0921         -7.68(*)
Liquidity (X7)                             0.0056          1.80
Variance of Earnings (X8)                 -0.7804         -1.23
Summary of Test Statistics
Durbin `h' (t-statistic)                  -0.0262         -1.17
Chow Test [[chi square](12)]              23.3185(*)

                                           Dependent Variables(2)
Independent Variables                          Long-Term Debt
                                         Estimate    t-statistic(1)

Innovations
Unexpected Investment (I1)                 0.0624        3.93(*)
Unexpected Retained Profit (/2)           -0.3910       -3.71(*)

Adjustment Mechanism
Change in Equity ([Delta]YE)               0.6789         2.64(*)
Change in Long-Term Debt ([Delta]YL)      -2.4976       -13.15(*)
Long-Run Variables
Interest Tax Shield (X1)                  0.0265          0.94
Non-Debt Tax Shield (X2)                  0.0002          0.80
Profitability (X3)                       -0.5110         -3.92(*)
Growth Opportunities (X4)                 0.0763          3.40(*)
Collateral Value of Assets (X5)           0.0416          1.23
Firm Size (X6)                            0.0729          7.01(*)
Liquidity (X7)                           -0.0146         -5.38(*)
Variance of Earnings (X8)                 0.8969          1.64
Summary of Test Statistics
Durbin `h' (t-statistic)                 -0.1014         -4.04(*)
Chow Test [[chi square](12)]             13.02

Independent Variables                       Short-Term Debt YS
                                         Estimate    t-statistic(1)

Innovations
Unexpected Investment (I1)                 0.1987       12.93(*)
Unexpected Retained Profit (/2)           -0.4566       -4.47(*)

Adjustment Mechanism
Change in Equity ([Delta]YE)              3.1525        12.68(*)
Change in Long-Term Debt ([Delta]YL)      2.1110        11.48(*)
Long-Run Variables
Interest Tax Shield (X1)                  0.0031         0.12
Non-Debt Tax Shield (X2)                 -0.0001        -0.28
Profitability (X3)                       -0.5568        -4.41(*)
Growth Opportunities (X4)                 0.2696        12.39(*)
Collateral Value of Assets (X5)          -0.0162        -0.49
Firm Size (X6)                            0.0193         1.91
Liquidity (X7)                            0.0090         3.42(*)
Variance of Earnings (X8)                -0.1165        -0.22
Summary of Test Statistics
Durbin `h' (t-statistic)                 -0.1926        -7.79(*)
Chow Test [[chi square](12)]             26.8917(*)

Independent Variables                 [sub.L][[Beta].sub.j]
                                      -[sub.s][[Beta].sub.j]
                                          Chi-Square (1)

Innovations
Unexpected Investment (I1)
Unexpected Retained Profit (/2)

Adjustment Mechanism
Change in Equity ([Delta]YE)
Change in Long-Term Debt ([Delta]YL)
Long-Run Variables
Interest Tax Shield (X1)                      0.27
Non-Debt Tax Shield (X2)                      0.45
Profitability (X3)                            0.49
Growth Opportunities (X4)                    29.08(*)
Collateral Value of Assets (X5)               1.14
Firm Size (X6)                               10.44(*)
Liquidity (X7)                               29.67(*)
Variance of Earnings (X8)                     1.35
Summary of Test Statistics
Durbin `h' (t-statistic)
Chow Test [[chi square](12)]

Notes:

(1.) An (*) indicates significance at the 95% confidence level for a two-tail test.

(2.) See the Appendix for variable definitions, The variables have been transformed by subtracting out individual firm mean values.

The summary test statistics in table 1 provide mixed results, with the null hypothesis of structural stability not being rejected by the Chow test at the 99% confidence level for the equity and long-term debt equations, while it is narrowly rejected for the short-term debt equation. Moreover, there is evidence from Durbin's h-statistic of first-order autoregressive serial correlation in the debt equations, though not in the equity equation.(16) Thus the model appears to perform better over time in explaining the choice between equity and debt than the choice between short- and long-term debt.

One indicator of the reasonableness of the model is the overall significance of the explanatory variables. In this respect the model performs quite well with all the innovations and adjustment mechanism parameters being statistically significant at the 90% confidence level (and 11 of the 12 at the 95% level) while half of the long-run parameters are also significant at the 90% level.

4.1 Long-Run Properties

Turning in table 1 to the long-run parameter estimates, [[Beta].sub.j], to facilitate the analysis of the parameters relating to maturity choice, the chi-squared statistic for [sub.L][[Beta].sub.j] - [sub.S][[Beta].sub.j] = 0 is shown in the far right hand column of the table. Of the eight hypothesised long-run determinants of capital structure, only four are statistically significant determinants of equity ratios and debt maturity. They are expected profitability, growth opportunities, firm size and liquidity.

Expected profitability is directly related to equity ratios with the offsetting effect on debt ratios evenly spread between short- and long-term debt so there is an insignificant effect on the difference between long- and short-term debt ratios. The equity result is consistent with the information asymmetry framework of Myers and Majluf (1984) which predicts a positive relationship between profit retention and the equity ratio but is inconsistent with the signalling models which predict an inverse relationship (see Blazenko). Moreover, we find no support for the signalling models of debt maturity as developed by Flannery, and Kale and Noe. While the maturity findings mirror the USA results of Barclay and Smith (1993), these authors correctly note that tests of signalling models using balance sheet data are not very powerful because the signalling models rely on `a transient state of informational asymmetry between managers and investors' (Guedes & Opler 1994).

The contracting cost hypotheses receive support from the significance of growth opportunities, firm size and liquidity variables. Growth opportunities, proxied by growth in total assets, are inversely related to both equity ratios and debt maturities. The latter is consistent with high growth potential firms with large information asymmetries shortening the maturity of their debt so as to reduce the under-investment problem. It also corresponds with the USA findings of Barclay and Smith (1993), and Guedes and Opler (1994). However, the inverse relationship with equity ratios is contrary to the traditional formulation of the underinvestment hypothesis where the substitution of equity for debt alleviates the under-investment problem. On the other hand, it is consistent with the Copeland and Weston (p. 507) argument that the dynamics of the pecking order theory implies that profitable firms in industries with few investment opportunities will end up with high equity ratios.

While the insignificance of the collateral value of assets variable is disappointing, the proxy for liquidity, the firm's current ratio, is likely to be inversely related to the collateralisable assets ratio. In this respect our results indicate that relatively liquid firms (with small collateralisable assets ratios) have significantly higher equity ratios and debt with shorter maturity, than less liquid firms. Hence, they are consistent with the USA results of Bradley, Jarrell and Kim (1984) for equity ratios and of Barclay and Smith (1993), and Guedes and Opler (1994) for debt maturity. Thus the evidence is consistent with the arguments of Myers (1977) and Diamond that firms match debt and asset maturities to lower agency costs of debt or to limit the risk of a borrower being forced into a costly liquidation because refinancing is not available.

Firm size, which is an inverse proxy for firm risk, is found to be inversely related to equity ratios and directly related to debt maturity. These results are supportive of the view that lenders are concerned with the moral hazard problem which arises when high risk firms undertake excessively risky projects to counter high risk premiums on the risky debt. With such lender concerns average debt maturity will decline, and equity ratios will increase, as small high risk firms are screened out of the long-term debt market. The corresponding USA results are somewhat mixed with Crutchley and Hansen (1989) and Barclay and Smith (1993) finding similar results to those reported here while Titman and Wessels, and Guedes and Opler find positive rather than inverse relationships.

Finally neither of the tax shield variables are statistically significant in any of the regressions. In part this reflects the relatively low power of tests based on balance sheet data, but it may also reflect the use of an average tax rate rather than a marginal rate. While MacKie-Mason's (1990) USA incremental decision study has found that the marginal tax rate affects the incremental debt-equity decision, a more recent USA study of incremental debt maturity decisions by Guedes and Opler finds no support for the tax hypothesis that debt maturity is inversely related to the expected marginal tax rate.

4.2 Dynamics of Capital Structure

The dynamic properties of the model arise from the coefficients of the adjustment mechanism variables, [Delta]YE and [Delta]YL, and of the innovation variables, I1 and I2. A unique feature of our research is the application of the general or interrelated adjustment model, rather than the univariate Koyck model. The primary distinguishing characteristic of the multivariate model is that the adjustment of any one dependent variable to its desired level depends not only on how far that variable is away from its desired value but it also depends on what is happening to the other dependent variables. We refer to these influences as cross-adjustment effects. The interrelated adjustment model collapses to the Koyck model when the cross-adjustment terms are zero. However, a joint test of significance of these cross-adjustment coefficients in table 1, that is ([sub.E][[Lambda].sub.1] - [sub.E][[Lambda].sub.2]) and ([sub.L][[Lambda].sub.1] - [sub.L][[Lambda].sub.2]), indicates that they are statistically significant.(17) Thus our estimates provide support for the interrelated adjustment model and the hypothesis that equity and debt decisions are interrelated.

Consistent with Jalilvand and Harris, our model also allows a firm's investment to impact on the capital structure decision. The statistical significance of the investment innovation (I1) and expected investment (X4) variables in table 1 provides support for the view that investment and capital structure decisions are interrelated. Furthermore, consistent with the transactions cost based predictions in (12), a positive innovation in investment results in a statistically significant reduction in the equity ratio and increases in the long- and short-term debt ratios.

As hypothesised in equation 13, retained profit innovations are found to have a statistically significant positive impact on the equity ratio with offsetting, and approximately equal, effects on long- and short-term debt ratios respectively. Moreover, the coefficient of the retained profit innovation variable in the equity equation is not significantly different from unity. This suggests that firms do not offset the retained profit shock by adjusting other sources of equity (ie, private or public placements) within the initial period of the shock.

The dynamics of our model allow innovations in investment and retained profit, I1 and I2, to have different impacts on capital structure. With the assumption that investment and retained earnings are exogenous, the sources and uses of funds identity requires innovations in these elements to impact on capital structure, though the impact coefficients would differ in sign. This prediction is strongly supported by the data with innovations in retained earnings increasing equity and decreasing debt ratios while innovations in investment decrease equity and increase debt ratios. Moreover, the parameter estimates indicate that a given innovation in retained earnings has approximately three times the impact on debt and equity ratios as an identical innovation in investment.(18) Hence studies, such as Jalilvand and Harris, which impose an identical adjustment on investment and retained earnings innovations, would appear to be misspecified.

Another feature of our model is the decomposition of actual retained earnings and investment into their expected and unanticipated (innovation) components and the independent modelling of each component. This specification can be tested against the alternative of including only actual retained earnings and actual investment in the model. If the latter formulation were correct then the coefficients on the expected and unanticipated components would be equal. However, application of the chi-squared test strongly rejects the null hypothesis of equality in the equity and short-term debt equations.(19)

4.3 Dynamic Simulations 

Because the estimated adjustment parameters in table 1 involve complex non-linear functions of [sub.E][[Lambda].sub.0], [sub.E][[Lambda].sub.1], [sub.E][[Lambda].sub.2], [sub.L][[Lambda].sub.0], [sub.L][[Lambda].sub.1] and [sub.L][[Lambda].sub.2], and these elements cannot be individually identified, the nature of the adjustment process in the estimated model is obscure. To gain a better understanding of the adjustment process we performed several dynamic simulations with the estimated model. The four which we report are: (i) a transitory or one period increase of 0.06 in a firm's retained profits to total assets ratio; (ii) a permanent increase of 0.06 in this ratio; (iii) a transitory increase of 0.1 in a firm's investment (change in total assets) to total assets ratio; and (g) a permanent increase of 0.1 in this ratio.

In the first simulation the firm's unexpected retained profit to total assets ratio is increased by 0.06 in period t = 1 but returns to its former level in t = 2 and subsequent periods. However, because of the assumption that expectations are based on a no-change forecast, the expected retained profit to total assets ratio increases by 0.06 in period t = 2 but then returns to its former level in t = 3 and subsequent periods. The multiplier effects on equity and debt ratios are depicted in figure 2a.(20) The transitory increase in retained profits to total assets increases the equity ratio over the first two years of the simulation but, thereafter, the equity ratio returns slowly to its pre-shock level. Moreover, the impact on the equity ratio is fully offset by reductions in each of the long- and short-term debt ratios of similar magnitude.

[Figure 2 ILLUSTRATION OMITTED]

The second simulation of a permanent increase in the retained profits to total assets ratio involves a single period increase in the unexpected retained profits ratio of 0.06 in t = 1, while expected retained profits is permanently increased by 0.06 from period t = 2. The multiplier effects on the three dependent variables are depicted in figure 2b. Adjustment to the new desired financial structure is shown to be quite slow with less than 80% of the adjustment being completed six years after the imposed shock to the model. Furthermore, the increase in the desired equity ratio is offset by similar reductions in both desired long- and short-term debt ratios, as was the case for transitory shocks in figure 2a.

Corresponding simulation results for transitory and permanent increases in investment are shown in figures 3a and 3b respectively. The major difference in the adjustment process for investment vis-a-vis that of retained earnings is the reduced role of long-term debt in the adjustment process for changes in investment. Thus the major effects of an increase in the investment to total assets ratio are reductions in the equity ratio and increases in the short-term debt ratio. Again the adjustment process is relatively slow being less than 80% complete after six years, suggesting the presence of significant adjustment costs in altering firms' financial structures.

[Figure 3 ILLUSTRATION OMITTED]

5. Summary and Concluding Remarks

The primary objective of the research reported in this paper was to enhance our understanding of the dynamics of financial structure decisions through the specification and estimation of a dynamic model of capital structure choice. A secondary objective was to test hypotheses concerning the long-run determinants of leverage and debt maturity within the Australian context.

In achieving these objectives, the study has developed a new methodological approach to the study of interrelated equity and debt financing decisions within a dynamic framework. This involved a simultaneous equation model with adding up constraints across the equity and debt equations. Moreover, we have estimated a multivariate adjustment model, rather than the simpler Koyck model, thereby incorporating an interrelated adjustment process whereby adjustment in any one liability category is influenced by changes in the remaining liabilities. Furthermore, our model recognises the interdependence of investment and financing decisions by incorporating exogenously determined innovations in investment and retained earnings as short-run determinants of capital structure.

A simulation analysis and tests of the dynamics of the econometric model revealed five primary features of the adjustment process of the capital structure of Australian firms. First, capital structure decisions are interrelated and need to be modelled using the multi-variate interrelated adjustment model. Second, investment and capital structure decisions are interrelated. Third, innovations in investment and retained earnings have different impacts on capital structure, so that prior studies which have constrained the effects of these innovations to be equal and opposite involve a misspecification of the adjustment process. Fourth, a formulation which decomposes actual investment and retained earnings into expected and unexpected components is superior to the alternative of including only the actual values of these variables in the model. Finally, the speed of adjustment to desired levels is quite slow with less than 80% of the adjustment being completed after six years, casting doubt on the validity of averaging three to five years of data as a means of netting out short-run disequilibrium in cross-sectional studies.

Our results are also of relevance to the debate on the determinants of a firm's optimal (or desired) capital structure. Broadly, the results mirror previous USA studies with the equity ratio positively related to expected profitability and inversely related to growth opportunities and firm size (an inverse proxy for firm risk). The results for profitability and growth opportunities lend support to the pecking order theory while the firm size effect is consistent with a contracting cost explanation with small high risk firms being limited in their access to debt markets. Furthermore, consistent with the contracting cost explanations of debt maturity choice small firms with extensive growth opportunities and relatively high liquidity tend to make greater use of short-term debt markets than long-term markets. The moral hazard problem associated with small (high risk) firms results in their being screened out of long-term debt markets, information asymmetries in firms with extensive growth opportunities leads to shortening of debt maturity so as to limit the under-investment problem, and agency costs of debt lead to a matching of debt maturities to asset maturities.

On the other hand, no empirical support was found for tax based explanations of capital structure decisions. However, our tests are based on balance sheet data and average tax rates and have relatively low power because the data reflects financing decisions over many prior periods. In light of this deficiency, it is desirable that future Australian capital structure research develop data bases of incremental capital structure decisions on which more powerful tests can be developed.

A difficulty with most capital structure studies is the need to proxy important theoretical determinants of capital structure choice. Titman and Wessels note that the use of such proxies generates numerous econometric problems including biased t-statistics, errors in variable problems, and spurious correlations. Hence our results need to be interpreted with considerable caution. As a means of mitigating measurement problems Titman and Wessels adopt the linear structural modelling approach within a static framework and the approach was subsequently applied to Australian data by Chiarella et al. (1992). As this research suffers from its neglect of the dynamics and interdependence of capital structure choice, future capital structure research could benefit from the incorporation of elements of the linear structural modelling approach within the dynamic framework developed in this study.

(Date of receipt of final typescript: July 1996. Accepted by Justin Wood, Area Editor.)

We acknowledge helpful comments and suggestions on earlier drafts of this paper from Neil Esho, Warren Hogan, Toan Pham, two anonymous referees, and participants at the Third International Conference on Asian-Pacific Financial Markets.

(1.) Essentially each determinant of the optimal or long-run capital (debt) ratio appears linearly and in multiplicative form with each of the variables influencing the adjustment speed (see Sharpe & Pooley 1991, p. 106).

(2.) This feature results from each of the liability (asset) equations having an identical set of explanatory variables with one of these variables being total liabilities (assets).

(3) Caution is required in evaluating this data as: (i) it relates only to book values as data for the market value of a firm's debt is not available; (ii) it relates to `gross' debt and does not net out holdings of financial assets by firms; (iii) the distinction between long-term and short-term debt is somewhat arbitrary with firms often classifying bank bill facilities of longer than twelve months as long-term liabilities; (iv) the data excludes financial leases which are a substitute for borrowings; and (v) the data includes non-interest bearing debt as well as interest bearing debt.

(4.) A fixed effects regression result with the equity ratio as dependent variable and a time frond as regressor produced a significant coefficient of the time trend variable.

(5.) Ryding (1990) provides a graph of the ratio of short-term to total debt for USA non-financial corporations based on Flow of Funds data. This data suggests that the ratio rose sharply from 1964 through 1982 but at its peak was less than 0.5.

(6.) Short-term liabilities may perform a buffer or residual role in liability management. However, as noted below, because the model is formulated with an identical set of explanatory variables in each of the behavioural equations, estimates of the model are invariant to the choice of residual liability.

(7.) For a discussion of liability (asset) interdependence sec Brainard and Tobin (1968) and Sharpe (1973).

(8.) In the equity equation this involves multiplying both sides by [[[sub.E][[Lambda].sub.0]/(1 + [sub.E][[Lambda].sub.0] [sub.E] [[Lambda].sub.2])].sup.-1] and then adding [YE.sub.it] to both sides and re-arranging the expression.

(9.) As in equations 6 and 7, the generalised adjustment parameters [sub.E][[Lambda].sub.0], [sub.E][[Lambda].sub.1], [sub.E][[Lambda].sub.2], [sub.L][[Lambda].sub.0][sub.L][[Lambda].sub.1] and [sub.L][[Lambda].sub.2] cannot be individually identified from regression estimates of equations 8 and 9.

(10.) This assumption implies a one way causal relationship with investment and dividend decisions affecting the financing decision. A more complete model would endogenise investment and dividend decisions by adding a further two equations to the model.

(11.) A referee has expressed concern that there appears to be an approximate identity between the variables in equation 8 as retained earnings cause an increase in equity as well as total assets. However, because total assets are exogenous in the model, retained earnings must be offset by either reducing long-term and/or short-term debt or by issuing less new equity. If the latter occurs, then the coefficient on the retained earnings variables in the equity equation will be less than unity.

(12.) The 1966 observation for each company is lost when lagged values are used.

(13.) Our data is ill ratio form, with each variable expressed as a ratio of total assets or profitability. The simple no change forecast is plausible for data in ratio form where the strong trend growth has been eliminated.

(14.) The fixed effects model does not allow for contemporaneous correlation in the regression error term across firms. Moreover, it restricts all but the constant term in the parameter set to be equal across firms. In contrast, the error components model restricts all parameter estimates to be the same across firms while the disturbance term is assumed to have two elements--a firm-specific random element and one that is random both across firms and over time.

(15.) A referee has noted that [YE.sub.i,t-1] and [YL.sub.i.t-1] may not be good instruments for [Delta] [YE.sub.i,t] and [Delta] [YE.sub.i,t]. However the use of instruments other than the lagged dependent variables will result in the estimates of the long-run coefficients and the adjustment process from equations 8 and 9 being different to those obtained from OLS estimation of equations 6 and 7 (which are mathematically equivalent to equations 8 and 9).

(16.) Because of the adding-up constraint across the three equations, any adjustment made for serial correlations must be made uniformly across the three equations.

(17.) The relevant chi-squared statistic with two degrees of freedom was 12.66.

(18.) A joint test that [sub.E][[Lambda].sub.3] + [sub.E][[Lambda].sub.4] and [sub.L][[Lambda].sub.3] + [sub.L][[Lambda].sub.4] are equal to zero is strongly rejected with a chi-squared statistic of 26.52.

(19.) For a joint test involving both retained earnings and investment the relevant chi-squared statistic with two degrees of freedom was 53.07 for equity and 58.90 for short-term debt. In the long-term debt equation the joint null hypothesis of [sub.L][[Beta].sub.4] - [sub.L][[Lambda].sub.3] = 0 and [sub.L][[Beta].sub.1] - [sub.E][[Lambda].sub.4] = 0 cannot be rejected, producing a chi-squared statistic of 1.74.

(20.) The multiplier is defined as the dynamic simulated value of the dependent variable with the transitory change in the retained profits ratio incorporated less the dynamic simulated value of the dependent variable when the retained profits ratio takes on historical values.

References

Barclay, M.J. & Smith Jr., C.W. 1993, The maturity structure of corporate debt, Unpublished manuscript, University of Rochester.

Barclay, M.J. & Smith Jr., C.W. 1995, The priority structure of corporate liabilities, Unpublished manuscript, University of Rochester.

Bewley, R. & Fiebig, D.G. 1990, `Why are long-run parameter estimates so disparate?', Review of Economics and Statistics, vol. 72, pp. 345-349.

Blazenko, G.W. 1987, `Managerial preference, asymmetric information, and financial structure', Journal of Finance, vol. 42, pp. 839-862.

Bradley, M., Jarrell, G.A. & Kim, E.H. 1984, `On the existence of an optimal capital structure: Theory and evidence', Journal of Finance, vol. 39, pp. 857-878.

Brainard, W. & Tobin, J. 1968, `Pitfalls in financial model building', American Economic Review, vol. 58, pp. 99-122.

Brick, I. & Ravid, A. 1985, `On the relevance of debt maturity structure', Journal of Finance, vol. 40, pp. 1423-1437.

Chiarella, C., Pham, T.M., Sim, A.B. & Tan, M.L. 1992, `Determinants of corporate capital structure: Australian evidence', in Pacific-Basin Capital Markets Research, vol. 3, eds S.G. Rhee & R.P. Chang, Elsevier Science Publishers, Amsterdam.

Copeland, T.E. & Weston, J.F. 1988, Financial Theory and Corporate Policy, 3rd edn, Addison-Wesley, Reading, MA.

Crutchley, C.E. & Hansen, R.S. 1989, `A test of the Agency Theory of managerial ownership, corporate leverage and corporate dividends', Financial Management, vol. 18, Winter, pp. 36-46.

De Angelo, H. & Masulis, R.W. 1980, `Optimal capital structure under corporate and personal taxation', Journal of Financial Economics, vol. 8, pp. 3-29.

Diamond, D.W. 1991, `Debt maturity structure and liquidity risk', Quarterly Journal of Economics, vol. 106, pp. 709-737.

Fischer, E.O., Heinkel, R. & Zechner, J. 1989, `Dynamic capital structure choice: Theory and tests', Journal of Finance, vol. 44, pp. 19-40.

Flannery, M.J. 1986, `Asymmetric information and risky debt maturity choice', Journal of Finance, vol. 41, pp. 19-37.

Guedes, J. & Opler, T. 1994, The determinants of the maturity of new corporate debt issues, Unpublished manuscript, Southern Methodist University.

Jalilvand, A. & Harris, R.S. 1984, `Corporate behaviour in adjusting to capital structure and dividend targets: An econometric study', Journal of Finance, vol. 39, pp. 127-145.

Judge, G., Griffiths, W., Hill, R., Lutkepohl, H. & Lee, T. 1985, The Theory and Practice of Econometrics, 2nd edn, Wiley, New York.

Kale, J.R. & Noe, T.H. 1990, `Risky debt maturity choice in a sequential equilibrium', Journal of Financial Research, vol. 13, pp. 155-165.

MacKie-Mason, J. 1990, `Do taxes affect corporate financing decisions?', Journal of Finance, vol. 45, pp. 1471-1493.

Marcus, A.J., 1983, `The bank capital decision: A time series--Cross section analysis', Journal of Finance, vol. 38, pp. 1217-1232.

Miller, M. 1977, `Debt and taxes', Journal of Finance, vol. 32, pp. 261-276.

Mitchell, K. 1991, `The call, sinking fund, and term-to-maturity features of corporale bonds: An empirical investigation', Journal of Financial and Quantitative Analysis, vol. 26. pp. 201-222.

Myers, S.C. 1977, `Determinants of corporate borrowing', Journal of Financial Economics, vol. 5, pp. 147-175.

Myers, S.C. 1984, `The capital structure puzzle', Journal of Finance, vol. 39, pp. 575-592.

Myers, S.C. & Majluf, N.S. 1984, `Corporate financing and investment decisions when firms have information that investors do not have', Journal of Financial Economics, vol. 13, pp. 187-221.

Nadiri, M.J. & Rosen, S. 1969, `Interrelated factor demand functions', American Economic Review, vol. 59, pp. 457-471.

Ryding, P., 1990, `The rise in U.S. corporate leveraging in the 1980s', in Studies on Corporate Leverage, ed. E.J. Frydl, Federal Reserve Bank of New York, New York, pp. 4584.

Sharpe, I.G. 1973, `A quarterly econometric model of portfolio choice: Part 1, Specification and estimation problems', Economic Record, vol. 49, December, pp. 518-533.

Sharpe, I.G. 1974, `A quarterly econometric model of portfolio choice: Part 2, The portfolio behaviour of Australian savings banks', Economic Record, vol. 50, March, pp. 21-38.

Sharpe, I.G. 1991a, `Deregulation, ownership structure and capital management of Australian savings banks 1967-1988', in Contemporary Issues in Australian Economics, eds M. Johnson, P. Kriesler & A.D. Owen., Macmillan, Melbourne, pp. 24-48.

Sharpe, I.G. 1991b, `Building society ownership structure and capital management', Applied Financial Economics, vol. 1, pp. 71-78.

Sharpe, I.G. & Pooley, G. 1991, `Asymmetric information, adjustment costs and capital structure of Australian finance companies', Pacific Accounting Review, vol. 3, June, pp. 93-125.

Taggart, R.A. 1977, `A model of corporate financing decisions', Journal of Finance, vol. 32, pp. 1467-1500.

Titman, S. & Wessels, R. 1988, `The determinants of capital structure choice', Journal of Finance, vol. 43, pp. 1-19.

Appendix

This appendix contains definitions of the dependent and independent variables used in the empirical study. With one exception, the M3 definition of the Australian money supply, the data for the 164 publicly listed companies was obtained from the AGSM Annual Report File. A listing of the companies and their Australian Stock Exchange Code may be obtained from the authors.

Dependent Variables

YE: The Equity Ratio--The ratio of equity to total assets where equity includes total shareholders funds (share capital, reserves and retained profits) and minority interests. YL: The Long-Term Debt Ratio--The ratio of long-term (deferred) liabilities to total assets where long-term liabilities include debentures, hire purchase payables, unsecured loans, mortgages and secured loans, amounts owed to associated companies, provision for deferred income tax and long service leave, and other deferred liabilities.

YS: The Short-Term Debt Ratio--The ratio of total current liabilities to total assets where total current liabilities include bank overdrafts, dividends and income tax payable, instalments on debentures/loans/mortgages and current loans, money accepted on deposit, trade creditors, accrued expenses and other provisions, and other current liabilities.

Long-Run Determinants of Capital Structure

X1: Expected Interest Tax Shield--Because of data unavailability we focus on the impact of the firm's expected effective tax rate defined as the lagged value of the ratio of reported company tax, TAX, to before tax income, NI, with negative values and values greater than unity being constrained as in Fischer, Heinkel and Zechner to a zero value for empirical purposes to preclude economically unreasonable effective tax rates:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

X2: Expected Non-Debt Tax Shield--Following Titman and Wessels a `direct estimate' of the non-debt tax shield is derived as:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

is the non-debt tax shield, NIBTID is net income before tax, interest and depreciation expenses, INT is interest expense, TAX is the reported company tax paid, and TAXR is the statutory corporate tax rate at time t. Although Titman and Wessels do not appear to have constrained the value of their non-debt tax shield variable, we have imposed the constraints to ensure the NDT variable takes on economically reasonable values.

X3: Expected Retained Profits Ratio--The lagged ratio of net income after tax and dividends, NIATD, to total assets.

X4: Expected Growth Opportunities--The lagged ratio of investment proxied by the change in total assets to total assets tie, the growth of total assets).

X5: Expected Collateralisable Assets Ratio--The lagged ratio of collateralisable assets, proxied by inventories and fixed assets, to total assets.

X6: Expected Firm Size--The lagged natural logarithm of the ratio of total assets to the M3 definition of the Australian money supply. Because the empirical analysis involves pooled time series and cross-sectional data it is necessary to scale total assets by a stock variable which reflects the general growth in the economy over-time.

X7: Expected Liquidity--The firm's lagged current ratio where current assets include cash on hand, money on deposit, marketable securities at market value, accounts receivable and inventories, while current liabilities include bank overdrafts, dividends payable, tax payable, instalments on debentures, money accepted on deposit and trade creditors. X8: Volatility of Earnings--Proxied by the variance of the annual change in the firm's net income before tax to total assets ratio over a moving five year period ending the current year. As an alternative proxy we experimented with the firm's systematic risk, or beta. A difficulty with this alternative was that almost half the sample companies had incomplete share price data, thereby restricting the sample size of the study.

Innovations

I1: Unexpected Investment--Proxied by the ratio of unexpected change in total assets to lagged total assets where the unexpected change is given by the change in the change of total assets (the second difference).

I2: Unexpected Retained Earnings--The ratio of the change in net income after taxes and dividends to lagged total assets.

Paul Gatward, Ian G. Sharpe, School of Banking and Finance, The University of New South Wales, Sydney NSW 2052; E-mail: [e-mail]