別訪問我！我對政治沒興趣：主題效應與TEDS高估投票率之研究

Abstract

「主題效應」(topic effect)指出，對調查主題越感興趣的人越可能接受訪問，所以和主題有關的估計值容易被誇大。台灣選舉與民主化調查(TEDS)是一個以政治及選舉爲主題的訪問，而投票與否這個問項又和這些主題息息相關，因此主題效應很可能是TEDS高估投票率的一個原因。本文用兩個假設來檢驗這個想法，首先在總體層次的分析上，本文發現TEDS高估投票率的情況比台灣社會變遷基本調查(TSCS)的更嚴重。由於兩個調查在主題外有許多相似性，這個結果表示兩者在投票率估計值上的差異很可能是主題效應所致。其次在個體層次的分析上，本文也發現在TEDS的成功樣本裡高投票意願者比低投票意願者更可能接受TEDS的訪問，符合主題效應的論述。整體來說，總體和個體層次的分析都支持本文的看法。不過，本文的結論是立基在特定前提之上，未來TEDS應蒐集更多資料來進一步確認及處理主題效應和高估投票率的問題。

Literature of ”topic effect” suggests that people who are interested in the survey topic are more willing to participate in the interview, so the estimates of the survey items that are associated with the topic tend to be exaggerated. Given that politics and election are the focal points of Taiwan`s Election and Democratization Study (TEDS), this paper explores whether the topic effect is one of the major causes to overestimate turnout rates in TEDS. The finding of the aggregate-level analysis shows that the turnout overestimate in TEDS is more serious than that in Taiwan Social Change Survey (TSCS). Since two survey projects have a lot in common except topics, their difference in turnout estimates may result from the topic effect. The individual-level analysis also indicates that the correlation between respondents` propensity to vote and their propensity to participate in TEDS is positive, which conforms to the theoretical expectation of topic effect. Generally speaking, the hypotheses are confirmed by both analyses. However, the conclusions are based on some untested assumptions; therefore future TEDS could use special survey design to further check and cope with the topic effect and turnout overestimate.