杜邦方程式三因子對半導體公司股價報酬的影響─動態縱橫資料模型之應用

文獻在估計股票報酬時，大多忽略股票報酬的遞延效果，以及ROE及杜邦方程式三因子（稅後淨利率、資產周轉率、權益乘數）應用於不同次產業公司將獲得不同的結果之事實。有鑑於此，本研究建立一個動態縱橫資料股價評估模型，以ROE的三個驅動因子及前一期股票報酬為解釋變數，評估該四個因素對股票報酬的影響。在估計上，將內生性問題納入考量，故採用一般動差法。實證上，以台灣上市櫃76家半導體公司為對象，研究期間為2008年第一季至2018年第二季，共3192筆觀察值。此外，為了評估不同次產業在結果上的差異，進一步將半導體產業區分為43家上游公司及33家中、下游公司。實證結果顯示：一、前一期股價報酬對當期股價報酬的影響，無論是全體樣本公司、上游公司或中、下游公司均是顯著的。因此，採用靜態縱橫資料模型進行估計，將產生偏誤，而利用動態追蹤資料模型估計杜邦方程式三因子對股價報酬的影響，可獲得更精確的估計結果。二、在動態縱橫資料模型下，對整體半導體樣本公司而言，ROE三個驅動因子對股價報酬的影響，除權益乘數不顯著外，稅後淨利率及資產周轉率皆為正向顯著的，且又以資產周轉率影響最大。換言之，半導體公司的股價實際受到資產使用效率的驅動較為明顯。三、將半導體公司進一步區分為上游公司與中、下游公司兩個不同的群體時，可以發現兩者的股價驅動因子存在明顯的差異。對中、下游半導體公司而言，資產周轉率對股價產生正向顯著的影響，且較全體半導體公司更為明顯，而上游半導體公司則產生顯著負向的影響。因此，即使在同一產業中， ROE三個驅動因子對股價報酬的影響將因上、中、下游公司而不同，若能將同產業進一步細分為不同的次產業後再行估計，會提供更精細的結果。四、無論是整體樣本公司、上游公司或中、下游公司，權益乘數對股價報酬的影響均是不顯著的，顯示財務槓桿的運用對半導體公司股價的影響不具顯著的效果。根據上述實證結果，本研究可提供業者及投資人改善獲利及篩選投資標的之依據。包括：一、投資人在利用杜邦方程式三因子以評估股價報酬時應加入前一期股價報酬率，亦即以動態縱橫資料模型進行估計，以免錯估其影響效果。二、在評估半導體公司股價報酬時，投資人宜進一步將半導體產業區分為上、中、下游產業，並分別對其進行估計，以獲得更精細且差異性的結論與投資決策。三、在杜邦三因子中，對半導體公司股價報酬的主要驅動因子是資產周轉率，其中上游公司為顯著負向的影響，中、下游公司則為顯著正向的影響。因此，當中、下游半導體公司更積極運用其資產時，應投資該等公司。中、下游公司業者則應強化資產使用效率，以增加淨利率，並獲取投資人認同，提升股價報酬，進而有利於利用股市籌措資金。四、無論是上游公司或中、下游公司，設法提升非來自於資產週轉與財務槓桿的稅後淨利率（例如：研發新產品、開拓新技術與新市場等），均有利於提升股價報酬，並利於籌措資金。

關鍵字

杜邦方程式；遞延效果；動態縱橫資料模型；一般動差法

並列摘要

When estimating stock returns, the literature mostly ignores the deferred effect of stock returns, and the fact that ROE and the three factors (net profit margin, asset turnover, equity multiplier) in DuPont equation will generate different results as the estimation is applied to different sub-industry. In view of this, this study establishes a dynamic panel data stock valuation model, using the three driving factors of ROE and the one-period lagged stock returns as explanatory variables to evaluate the impact of these four factors on stock returns. In the estimation, the generalized method of movement is adopted to resolve the endogenous problem. Empirically, a panel data set of 76 semiconductor companies listed in Taiwan during 2008: Q1 to 2018: Q2 is used, with a total of 3,192 observations. To assess the difference in results between different sub-sectors, the semiconductor industry is further divided into 43 upstream companies and 33 midstream and downstream companies. The empirical results show that: First, the impact of the previous stock returns on the current stock returns is significant for all sample companies, upstream companies or midstream and downstream companies. Thus, using a static panel data model to estimate the stock returns will produce bias. In other words, using the dynamic panel data model to evaluate the influence of the three factors in DuPont equation on stock returns can obtain more accurate results. Second, under the dynamic panel data model, for the whole semiconductor sample company, the impact of the equity multiplier on the stock returns is insignificant, and the impacts of the net profit margin and asset turnover are positive and significant. Moreover, the asset turnover has the greatest impact on the stock returns. Evidently, the semiconductor’s stock returns are mainly driven by the efficiency of asset use. Third, the impacts of the three driving factors of ROE on the stock returns vary with different semiconductor sub-industries. Asset turnover has a significant positive impact on stock returns for midstream and downstream semiconductor companies and has a significant negative impact for upstream semiconductor companies. Obviously, if a same industry can be further divided into different sub-industries, more detailed estimation results can be obtained. Fourth, for the overall sample company, upstream companies or midstream and downstream companies, the impact of equity multiplier on stock returns is insignificant, indicating that the use of financial leverage has no significant effect on the share returns. Based on the above empirical results, this study provides the following suggestions for the companies and investors to improve profitability and screen investment targets. First, investors who use the DuPont equation to evaluate stock returns should add to the one-period lagged stock returns as the regressor, that is, using a dynamic panel data model so as not to misjudge the impacts of the three driving factors of ROE on the stock returns. Second, in assessing the stock returns of semiconductor companies, investors should divide them into upstream, midstream, and downstream companies, and then can obtain more detailed and differentiated conclusions and investment decisions. Third, the main driving factor for the semiconductor company's stock returns is the asset turnover. However, for the upstream companies, the impact of the asset turnover is significantly negative, and for the midstream and downstream companies the impact is significantly positive. Thus, when midstream and downstream companies are more active in using their assets, they should invest in such companies. In addition, midstream and downstream companies should strengthen asset use efficiency to increase net profit margin and obtain investor recognition, improve stock returns, which will be helpful for the companies to raise funds in stock markets. Finally, for upstream companies and midstream and downstream companies, improving the net profit margin through the developments of new products, new technologies, or new markets are beneficial to raise stock returns and funds.