近年來台灣地區由於婦女受高等教育比例、就業率的增加等因素，使得育齡婦女的生育時間延後、生育總胎次減少，因此總生育率也逐年降低，在2001 年時降至1.4 的歷史新低，加速台灣地區人口老化的腳步。台灣地區婦女生育率的下降，在各年齡層有不同的趨勢，這種各年齡層生育率變化不同步的現象，使得許多常見的生育率模型 (如Gamma 及Lee-Carter 模型) 套用於台灣地區的生育率時產生不小的誤差。為尋求更精確的台灣地區生育率模型，除了考量直接預測年齡別生育率的各種模型外，本文也嘗試以世代 (Cohort ) 的觀點預測未來的生育率，包含世代生育率模型，以及研究胎次別生育率與年齡別生育率之間的關係，再套用國外近年提出的擴散 (Diffusion) 模型。本文的實證研究以西元1949 至1996 年 (或1975 至1996 年，對胎次別資料而言) 的資料為基礎，1997 至2001 年資料為檢測樣本，藉由交叉驗證比較上述方法的優劣；此外，本文也考慮中國習俗對十二生肖的好惡，將龍虎年的效應列入預測方法的修正。研究發現世代模型可能因資料分析及估計的程序較為複雜，預測誤差反而高於單一年齡組個別估計法這類模型較為簡單的方法。

The total fertility rates in Taiwan area have been falling rapidly since 1960's, which reached a record low 1.4 in 2001, and trigger the population aging in Taiwan area. Possible causes attributing to this rapid decrease include more opportunities of higher education and employment for the female, which generally are treated as two of the main reasons for the delay of childbearing. However, the falling of fertility rates vary quite a lot in different age groups, and thus most well-known fertility models (such as Gamma and Lee-Carter models) fail to produce accurate predictions. In this paper, we shall extend the study of Huang and Yue (2002) and continue looking for acceptable fertility projection models. The models considered in this study include, in addition to traditional models for age-specific fertility rates, models for cohort fertility rate (CFR), age-and-parity specific birth rate, and diffusion. Data from 1949 to 1996 (or 1975 to 1996 for age-and-parity specific birth rate) are used as a pilot for verifying the model which has the best fit for data gathered from 1997 to 2001. We also consider the effect of Chinese zodiac to adjust the prediction values. We find that simpler models, such as individual group estimation, have smaller prediction errors comparing the complicated models such as the CFR model.