臺灣糧食自給率下滑之實證研究

Abstract

臺灣作為高度依賴進口的島國，其糧食自給率自民國70年代以來長期下滑，現已降至約三成，反映本地農業生產能力與糧食自主性的持續弱化。過去學界多將焦點分別置於飲食結構變遷、耕地變化、價格支持或貿易自由化單一面向，較少系統性整合上述結構性因素對糧食自給率的交互影響。現行官方熱量自給率指標因將進口原料加工產出納入計算，潛在高估本土農業之貢獻，極需提出更貼近實質生產結構的補充衡量方式。 本研究以民國73至112年臺灣年度資料為基礎，嘗試建構「調整熱量自給率」指標，藉由剔除進口原料加工熱量，試圖補足官方統計之高估偏誤。實證部分分別以官方與調整後自給率為應變數，納入小麥純糧食供給量、休耕轉作補貼金額，並控制稻米保價收購金額與貿易自由化（WTO）等關鍵變數，利用ARDL-ECM模型進行共整合與長期效果檢驗。 本研究針對官方與調整後熱量自給率分別建立三組主要模型（共六個ARDL-ECM模型），檢視不同政策與結構性因素之長期影響。實證結果顯示，小麥純糧食供給量每增加1公斤，官方自給率長期下降約0.287%至0.298%，調整後自給率則下降0.317%至0.408%，顯示飲食結構西化對本土自給力之負面衝擊在調整後指標下更加明顯。休耕轉作補貼金額每增加1億元，官方自給率長期減少0.011%至0.019%，調整後則減少0.014%至0.021%，突顯休耕政策對糧食安全的直接風險。農業補貼每增加1億元，熱量自給率於官方模型長期提升0.044%至0.048%，調整後則略為收斂至0.036%至0.039%，但於所有模型皆為顯著正向。至於WTO（貿易自由化）變數，僅在官方自給率模型下於加入當年呈現顯著負向衝擊（約-1.218%），調整後模型則因已將進口熱量內生化，WTO效果不再顯著。最後，六個模型中ECM誤差修正項均為顯著負值（約-0.42至-0.72），顯示當自給率偏離長期均衡時，系統具備高度動態調整能力，平均2至3年即可回復均衡狀態。

Taiwan, as an island nation highly dependent on imports, has experienced a long-term decline in its food self-sufficiency ratio since the 1980s, which has now dropped to about 30%. This trend reflects the persistent weakening of domestic agricultural production capacity and food autonomy. Previous academic studies have mostly focused on individual aspects such as dietary structure transformation, farmland changes, price support, or trade liberalization, with relatively few systematically integrating the interactive effects of these structural factors on the food self-sufficiency ratio. The current official calorie-based self-sufficiency indicator incorporates the output of imported raw material processing, potentially overestimating the contribution of domestic agriculture. Therefore, it is crucial to propose supplementary measures that more accurately reflect the actual production structure. Based on annual data from 1984 to 2023, this study attempts to construct an “adjusted calorie-based self-sufficiency ratio” by excluding the calorie contribution of imported raw material processing, aiming to correct the overestimation in official statistics. In the empirical analysis, both the official and adjusted self-sufficiency ratios are used as dependent variables. The main explanatory variables include wheat net food supply per capita and the amount of fallow and crop diversification subsidies, while the models control for rice price support (procurement subsidies) and trade liberalization (WTO membership). The ARDL-ECM model is used to examine cointegration and long-term effects. This study establishes three main models for both the official and adjusted calorie-based self-sufficiency ratios (a total of six ARDL-ECM models) to assess the long-term impacts of different policy and structural variables. The empirical results show that for every additional kilogram of wheat net food supply per capita, the official self-sufficiency ratio decreases by about 0.287% to 0.298% in the long run, while the adjusted self-sufficiency ratio declines by 0.317% to 0.408%, indicating that the negative impact of dietary westernization on domestic self-sufficiency is even more evident under the adjusted indicator. For every additional NT$100 million in fallow and crop diversification subsidies, the official self-sufficiency ratio decreases by 0.011% to 0.019% in the long run, and the adjusted ratio decreases by 0.014% to 0.021%, highlighting the direct risk that fallow policies pose to food security. Each additional NT$100 million in agricultural subsidies increases the official self-sufficiency ratio by 0.044% to 0.048% in the long run, while the adjusted ratio rises by 0.036% to 0.039%; both effects are significantly positive in all models. As for the WTO (trade liberalization) variable, a significant negative impact is observed only in the official self-sufficiency ratio model in the year of accession (about -1.218%), whereas the effect is no longer significant in the adjusted model due to the endogenization of import-related calories. Finally, in all six models, the ECM error correction terms are significantly negative (about -0.42 to -0.72), indicating a high degree of dynamic adjustment, with the system returning to equilibrium within an average of two to three years when the self-sufficiency ratio deviates from its long-term equilibrium.