影響台灣上市食品類股股價報酬率的因素 -ARDL/NARDL模型之應用

Abstract

本文研究目的在於分析台灣上市食品類股股價報酬率的影響因素，透過分析台灣上市食品股，篩選近五年平均 EPS 大於 3.0 元、ROE 高於 9%，且本益比介於 10 至 25 倍之公司，此標準旨在剔除極端估值與獲利波動過大者，以確保研究樣本具備穩健的財務基礎與資本運用效率。篩選後結果以台灣三家代表性食品公司為研究標的，統一企業、大成食品與卜蜂食品，選擇此三家公司主要基於三項考量，其一，三者在台灣食品供應鏈中占有高度市占與代表性，涵蓋食品製造、飼料、肉品加工等上中下游重要環節；其二，三家公司營運模式與成本結構具差異性，使其對總體環境與原物料價格變動的反應可能呈現不同樣貌，適合理解產業內部的異質性；其三，三者皆具備完整且穩定的公開財務資訊，可支撐長期時間序列分析之需求。研究期間以2004年1月至2024年12月之股價資料計算其股價報酬率，利用ARDL/NARDL模型分析三間公司之股價報酬率。 首先，經由單根檢定確認所選變數之整合階數適合進行自我迴歸分配遞延模型(autoregressive distributed lag model, ARDL)與非線性自迴歸分佈遞延模型(Nonlinear Autoregressive Distributed Lag, NARDL)分析。其後，採用ARDL–誤差修正模型(Error Correction Model, ECM)與邊界檢定驗證了股價報酬率與解釋變數之間穩定的長期均衡關係；並透過NARDL模型拆解利率與匯率的正負衝擊，以找出市場對政策利多與利空消息的結構性差異。 實證結果顯示，短期內，三家公司的股價均對原物料期貨價格與企業經營指標呈現明顯的多期落遲效應，尤以統一企業的成本反應最為顯著；名目GDP與月營收成長率的突發變動往往伴隨股價的逆向調整；ROE與毛利率改善則可即期或數月後提升股價回報。NARDL分析進一步發現，降息對統一與大成具有即期正面激勵，而升息抑制效應已在市場中提前反映；雖然台幣貶值通常會推升進口原物料成本，但實證結果顯示，台幣貶值反而與卜蜂與大成的股價報酬率呈現正向關係，顯示其具備一定的成本轉嫁能力；相較之下，台幣升值對股價的影響則不具一致性。 基於上述結果，建議投資人短線布局應重點關注期貨與匯率的交替震盪節奏以及企業獲利指標的落遲反應；長線持有則宜依循毛利率、ROE與營收持續成長所揭示的價值驅動；企業經營者應強化成本控管，並持續優化營運效率；政策制定者則可透過適時的貨幣與匯率指引，協助穩定食品供應鏈成本波動。未來研究可擴充至更高頻或更長期資料，並納入ESG、消費者行為等新變數，以提升模型預測與策略建議之完備性。

The objective of this study is to analyze the determinants of stock returns for listed food companies in Taiwan. The study screens for companies with a five-year average Earnings Per Share (EPS) greater than 3.0 TWD, a Return on Equity (ROE) exceeding 9%, and a Price-to-Earnings (P/E) ratio between 10 and 25. These criteria aim to exclude companies with extreme valuations or excessive earnings volatility, ensuring that the research sample possesses a robust financial foundation and capital efficiency. Based on these criteria, three representative companies—Uni-President Enterprises Corp., Dacheng Foods Co., Ltd., and Bofeng Enterprises—were selected as the research targets. This selection is based on three considerations: first, they hold significant market share and representativeness within Taiwan's food supply chain, covering key upstream, midstream, and downstream sectors such as food manufacturing, feed, and meat processing; second, their differences in business models and cost structures allow for an analysis of industry heterogeneity regarding responses to macroeconomic environment and raw material price fluctuations; third, they possess complete and stable public financial data suitable for long-term time series analysis. The study period covers stock price data from January 2004 to December 2024 to calculate stock returns, employing the Autoregressive Distributed Lag (ARDL) model and the Nonlinear Autoregressive Distributed Lag (NARDL) model for analysis. Methodologically, Unit Root Tests were first conducted to confirm that the integration orders of the selected variables were suitable for ARDL and NARDL modeling. Subsequently, the ARDL-Error Correction Model (ECM) and Bounds Test verified a stable long-term equilibrium relationship between stock returns and the explanatory variables. The NARDL model was then utilized to decompose the positive and negative shocks of interest rates and exchange rates, identifying structural differences in market reactions to favorable and unfavorable policy news. Empirical results indicate that, in the short term, the stock prices of all three companies exhibit significant multi-period lagged effects in response to raw material futures prices and corporate operating indicators, with Uni-President showing the most significant reaction to costs. Sudden fluctuations in nominal GDP and monthly revenue growth rates are often accompanied by inverse adjustments in stock prices, whereas improvements in ROE and gross margin can lift stock returns immediately or after a delay of several months. The NARDL analysis further reveals that interest rate cuts provide an immediate positive stimulus for Uni-President and Dacheng, while the suppressive effect of rate hikes appears to be priced in by the market beforehand. Although the depreciation of the New Taiwan Dollar (TWD) typically drives up imported raw material costs, the results show a positive correlation between TWD depreciation and the stock returns of Bofeng and Dacheng, suggesting they possess a certain degree of cost pass-through capability. In contrast, the impact of TWD appreciation on stock prices lacks consistency. Based on these findings, investors are advised to focus on the alternating rhythms of futures and exchange rates and the lagged reactions of corporate earnings indicators for short-term positioning; for long-term holding, strategies should follow value drivers such as sustained growth in gross margin, ROE, and revenue. Corporate managers should enhance cost control and continuously optimize operational efficiency. Policymakers can assist in stabilizing costs within the food supply chain through timely monetary and exchange rate guidance. Future research could expand to include higher-frequency data or a longer timeframe, and incorporate new variables such as ESG factors and consumer behavior to enhance the completeness of model predictions and strategic recommendations.