Determination of Capital Structure and Prediction of Corporate Financial Distress
Hussain, Mohmad Isa (2005) Determination of Capital Structure and Prediction of Corporate Financial Distress. PhD thesis, Universiti Putra Malaysia.
This study examines the determinants of capital structure of 182 Malaysian listed firms utilizing panel data from 1986-2001. To enhance the capital structure model, this study incorporated macroeconomics variables together with the traditional financial ratios in determining the capital structure choice. Besides, this study also employs the dynamic capital structure model, using panel data analysis, to estimate the parameters of interest and the speed of adjustment of Malaysian listed firms towards target level of leverage. In fact, this is the pioneer attempts in the application of the dynamic analysis to capital structure model and utilization of large data set of Malaysian listed firms. Thus the results would be of great contribution especially in the context of the emerging market. Empirical results show the following. First, the results of the static capital structure model using the pooled OLS estimation and Fixed Effects (FE) models were analyzed and compared. Of these static models, after correcting for heteroscedasticity and auto correlation problems, the Generalised Least Square (GLS) method is the best static model because it has the higher goodness of fit of 90.94% compared to 57.52% (i.e. comparison between the Lev6 of market value model of GLS estimation and the Lev6 of market value model estimated by Transformed Regression Model). Second, the dynamic capital structure model was estimated using a must stronger estimation technique, Generalised Method of Moments (GMM). Under the GMM estimation, this study deploys a consistent estimation method as suggested by Anderson and Hsiao (1982) and Arellano and Bond (1991). For comparison purposes, pooled OLS estimates were also obtained. After comparing the results, this study concluded that Arellano and Bond's method is the most appropriate for the dynamic model because the performance of its estimators results in smaller variances than those associated with Anderson and Hsiao's approach. The final dynamic capital structure model reveals that 13 variables were significantly related with the level of leverage and eight variables were not significant. In addition to firm-specific characteristics, this study found that macroeconomics variables are also important factor in determining the financing decision. These empirical findings support the study hypothesis that BPUET*AKAAN SULTAN A W L W wmm m MALAYSIA v macroeconomics factors were also important and would affect capital structure choice. The three most significant determinants are, (i) lagged leverage, (ii) nondebt tax shield, and (iii) money supply. The sign of these relations suggest that both the pecking order theory and the trade off theory are at work in explaining the capital structure. The results also show that Malaysian firms adjust toward target leverage but the speed of adjustment of 0.47 is slower compared to 0.57' for developed countries such as United States and United Kingdom. Besides, it seems that the cost of deviating from the target leverage is not generally large enough to motivate costly external capital market transaction. It was observed that the capital structure model reported in the literature especially for Malaysian has only been short-term in nature because they are based on a static snapshot framework. The empirical evidence of this study clearly indicates that the findings from such studies were found to be seriously underestimating the impact of the explanatory variables in the long-term equilibrium. This long-term outlook and its finding is a new contribution to the issue in the Malaysian context. In the second part, two corporate financial distress models were constructed for Malaysian listed firms. Eight independent variables were used for the capital structure prediction model (CS-prediction model), while nine selected literature based variables were deployed for literature based prediction model (L-prediction ' This is an average sped of adjustment for United States and United Kingdom. model) and observed the models' accuracy. The in-sample overall accuracy of the CS-prediction model is 71.1% and the L-prediction model is 85.2%. The Nagelkerke R~ of the CS-prediction is 45.50% while L-prediction model is 62.40%, which implies that relatively the literature based predictors of the model significantly explained the contribution to the financial distress. Further, the predictive power of both models was tested using the holdout samples. Comparatively, for the first three years period prior to distress, this study found that the L-model consistently outperformed the CS-model. In fact, the results of Lmodel demonstrated excellent Type I accuracy2 of 1 OO%, Type 11 accuracy3 of 90% and overall accuracy of 95.00% one year prior to distress. It was also observed that the overall accuracy remained high for the second year (94.99%) and the third year (84.99%). The estimation results of L-prediction model confirmed all the expectations. The model indicates that declining profit margin on sales (T1) and operating efficiency (T9) contributes significantly towards the firm becoming financially distressed, while the total debt ratio (T6) and current liabilities to total assets ratio (T7) are shown to have direct contribution to the financial distress. Of these significant variables, the total debt ratio (T6) and current liabilities to total assets ratio (T7) were found to be the two most significant factors in determining the outcomes of * Correctly classify a financially distressed firm as distressed firm. Correctly classify a healthy firm as healthy. financial distress with the largest elasticity value of 14.1600 and 10.3480 respectively. In general, these results are consistent with the trade-off theory which predicts that highly leveraged firm is vulnerable to financial distress. The results also shed some light on the factors that caused financial distress to many Malaysian listed companies. Following these results, the study concluded that firm with less profit margin on sales (TI) and operating efficiency (T9) and high in total debt ratio (T6) and current liabilities to total assets ratio (T7) would have higher financial distress prospect in Malaysia. In sum, the L-prediction model is the preferred corporate financial distress prediction model and is capable of providing effective early warnings information of financial distress especially three years period prior to distress.
Repository Staff Only: Edit item detail