浮誇新聞與高產榮景：以文字探勘分析大躍進的饑荒成因

本研究主要探討「大躍進」時期，不同地區的浮誇新聞發佈數量對該地區饑荒死亡人數之影響。從過去文獻可以得知，「大躍進」時期中央常下達遙不可及的高產量任務，導致地方幹部浮報產量的行為層出不窮。當時中國最大的官方媒體《人民日報》更大肆地對這種浮誇資訊加以宣傳，導致全國吹起了一陣「浮誇風」，社會充滿著大量虛假的浮誇新聞。浮誇訊息的傳播不僅讓中央領導者誤以為當時「大躍進」政策正往好的方向發展，進而逐步提高各地糧食生產與徵購的目標，而幹部為了完成因「浮誇風」不斷提高的糧食生產任務，開始不擇手段地從農民身上榨取糧食，中央更因有些地區無法達成生產目標而懷疑當地農民「瞞產私分」才導致缺糧，於是引發一系列官員暴力掠奪糧食的整風，最終各地開始爆發饑荒，大量人民的死亡反映了「浮誇風」的後果。本研究從《人民日報》官方資料庫擷取了1958年至1962年發佈的新聞報導，經過資料清理與文本相似性計算篩選出浮誇新聞文本。使用關鍵字分析這些浮誇新聞後發現，《人民日報》所報導的糧食浮誇新聞主要是利用「豐收」、「高產」、「破紀錄」等誇大的形容詞，以及「萬畝」、「萬斤」、「千萬噸」等巨量的單位來塑造當時社會繁華的盛況，其最終目的是為了激勵各地幹部的生產進度以及強化民眾對中央政策的認同感，使「大躍進」政策可以順利執行。得知浮誇新聞之定義與組成後，本研究透過字典搜索方法將大量浮誇新聞以省級行政區為單位進行分類，得出不同省份的浮誇新聞數量，藉以衡量當地「浮誇風」的氾濫程度。以「浮誇風的氾濫使中央誤信高產狀況而提高生產目標，導致一系列打壓農民、榨取糧食的行為發生，最終饑荒死亡人數增加」做為影響途徑，假設「浮誇新聞越多會導致饑荒死亡人數越多」。為了驗證該假設，本研究將以「大躍進」五年期間各省的「浮誇新聞發佈百分比」作為解釋變數，各省份「人均饑荒非正常死亡百分比」作為被解釋變數進行迴歸分析。研究發現「不同地區的浮誇新聞數量對饑荒非正常死亡人數有顯著的正向影響」然而有少數饑荒特別嚴重的地區，如四川、安徽則會因為其死亡人數過多而被模型嚴重低估，這表示模型較無法解釋饑荒特別嚴重的地區。

關鍵字

大躍進；大饑荒；文字探勘；中共官媒；迴歸分析

並列摘要

It can be seen from past study that during the "Great Leap Forward" period, the central government often require high-yield tasks, causing local officials to overstate their output. At that time, the People's Daily, China's largest official media, promoted this kind of exaggerated information more aggressively, causing a "wind of exaggeration" across the country, and society was full of a lot of fake exaggerated news. The exaggerated information not only made the central government believe that the "Great Leap Forward" policy was going well at that time, and gradually increased the goals of grain production and procurement in various regions, but also the officials were in order to complete their tasks, began to snatch food from farmers by irregular means. The central government even suspected that local farmers were "hiding food" because some areas could not achieve production goals, which led to the shortage of food. Then led to a series of rectification in which officials violently looted food, and eventually famines began to occur in various places, and the deaths of large numbers of people. This study finds reports published from 1958 to 1962 from the official database of People’s Daily, using data cleaning and text similarity screening, and finally classified food-related exaggerated news in different regions by provincial administrative regions. First of all, in order to understanding of the composition and structure of exaggerated news, we can see from the results of keyword analysis that the exaggerated grain news reported by People’s Daily mainly uses exaggerated adjectives such as “bumper harvest”, “high yield” and “record breaking”, and Exaggerated unit quantifiers to shape the prosperity of the society at that time. Through keyword analysis, we can understand the context and conceptual structure of exaggerated news. The purpose is to promote the production progress in various places, communist thought, and to strengthen the public's sense of identity with the central policy, so that the "Great Leap Forward" policy can be implemented. After understanding the definition and composition of exaggerated news, this study classified a large number of exaggerated news by provincial administrative regions through the dictionary search method, and obtained the number of exaggerated news in different provinces, so as to measure the prevalence of local "exaggerated wind”. Using exaggerated news to convince the central government to believe in high production conditions and increase production goals, a series of behaviors that hurt farmers, snatch food, and eventually increase the number of deaths from famine. To test this hypothesis, this study will use the percentage of exaggerated news releases in each province during the five-year period of the Great Leap Forward as the independent variable, and the per capita famine death rate per capita in each province as the dependent variable for regression analysis. The study found that the number of exaggerated news in different regions has a significant positive impact on famine deaths. However, there are a few regions with particularly severe famine, such as Sichuan and Anhui, which are seriously underestimated by the model because of their excessive number of deaths, which means that the model is less able to explain areas with severe famine.