Document Type
Article
Abstract
This study adopted multiple linear regression models and artificial neural networks (ANNs) to analyze the important determinants of capital structures of the high tech and traditional corporations in Taiwan, respectively. The ten independent variables (determinants) employed herein included seven corporation feature variables and three external macro-economic variables. The following conclusions were reached: 1) From the root MSE, the ANN model achieved a better fit than the regression model. 2) The capital structure of high tech corporations does not differ significantly from that of traditional corporations, but differences do exist in the determinants of the capital structure. 3) Macro-economic variables more significantly affect the sensitivity of the capital structure of high tech corporations than traditional corporations. 4) Business risk has positive/negative impacts on capital structure of high tech/traditional corporations, respectively. 5) Six features of corporations have the same impacts on both high tech and traditional corporations, namely: firm size (+), growth opportunities (+), profitability (-), collateral value (+), non-debt tax shield (-), and dividend policy (-). In optimizing capital structure, the following policy implications can be dra wn for any company based on the results of this study: l Larger corporations can borrow more than small corporations, and thus enjoy the benefit of greater financial leverage. l Corporations with higher growth opportunities need to borrow more to meet their capital needs. l Corporations with higher profitability need to borrow less to meet their capital needs. l Corporations with higher collateral value (fixed assets) can borrow more than those with lower collateral value. l Increased non-debt tax shield will lower the tax benefits of financial leverage and hence reduce incentives for borrowing. l Corporations with higher cash dividend payments generally borrow less than corporations with lower cash dividend payments. Managers can apply the analytical results above to optimize capital structure and maximize firm value.
Recommended Citation
Pao, Hsiaotien; Lee, Tenpao; and Tong, Leeing, "Estimating the Capital Structure of High Tech and Traditional Corporations' Capital Structure: Artificial Neural Networks vs. Multiple Linear Regressions" (2001). ICEB 2001 Proceedings (Hong Kong, SAR China). 114.
https://aisel.aisnet.org/iceb2001/114