台灣總生育率再分析－ACF法的運用

台灣總生育率(TFR)的持續下跌引發眾多討論，TFR為一種估計測量，容易受婦女生育的數量(quantum)與步調（tempo，又稱生育的時間效果）之變化而影響，由於許多國家調查的「理想子女數」仍保持在兩名水準，但是TFR卻遠低於替換水準，Bongaarts and Fenny (1998)認為這是生育步調延後所致，婦女實際的生育水準不若觀察值那樣低。不少學者提出調整方法，希望降低婦女生育步調的影響，以獲得實際的生育水準，其中，Bongaarts等人提出的B-F法運用最為廣泛，然而，B-F法只是去除生育時間效果後的總生育率調整值，依舊無法代表實際生育水準，必須以其他方法來探討女性的生育水準。我們藉由Schoen (2004)的ACF法來調整TFR，只是ACF法得待所有婦女完成生育方能運用，本文乃透過Li and Wu (2003)建議的SVD模型推估未完成生育之年輪未來的生育率。本文使用B-F法與ACF法來調整1980年至2005年台灣的TFR，經過比較，ACF法的結果比B-F法之估計更加接近CFR，二種結果都指出目前台灣婦女之生育水準仍高於千分之1300（2005年B-F法之結果為千分之1545，ACF法的結果為千分之1399），調整值雖未落入「超低生育水準(lowest-low)」，但也不若「理想子女數」高。

關鍵字

總生育率；生育步調； B-F法； SVD模型； ACF法

並列摘要

The TFR in Taiwan has declined more rapidly and pervasively than was expected. The TFR is a hypothetical measure, which can be influenced by two distinct parts: a 'quantum' component and a 'tempo' component. 'Many countries' 'desired children number' is kept at 2, but the TFR is far below 2. Bongaarts and Fenny criticized the TFR due to the changes in the timing of childbearing. Therefore, they announced the B-F method to shed the timing effect in order to more reliably capture fertility. However the tempo-free TFR method still can't represent the reality of fertility; we have to make use of other measures. We employed Li and Wu's SVD model and Schoen's ACF method to adjust the 1980-2005 TFR in Taiwan. The output of the ACF method was closer to the CFR. Both findings of the B-F method and ACF measure indicated that the TFR in Taiwan was above 1.3, the lowest-low fertility level. But the adjusted TFRs still did not reach 2.