| 摘要: | 先天因素和後天環境對子女教育成就的影響究竟孰重孰輕一直是經濟學與教育學界相當受到矚目的議題。Becker and Tomes (1986)和Haveman and Wolfe (1995)主張高教育程度的父母，透過人力資本的投資，小孩會有高的教育成就，這個觀點偏重於後天培養(nurture)對教育的重要性。Hernstein and Murray (1994)則持有不同的看法，認為高教育水準的父母能力(ability)好，智商(IQ)高，若這種能力透過遺傳基因傳遞給子女，則家族先天的良好基因(nature)會代代相傳給子女，子孫會持續擁有較高的教育成就。過去國外的文獻在討論這個議題時採用的實證策略大致可分為三種。第一種策略是使用包含同卵和異卵雙胞胎資料，在家庭成長環境相同的情形下，異卵雙胞胎和同卵雙胞胎後天表現的差異即可歸諸於遺傳基因上的不同(Plomin, DeFies and McClearn, 1990)。由於雙胞胎資料多為小樣本，同時同卵雙胞胎中兩個小孩被父母同等對待的機會大於異卵雙胞胎，因此，在估計先天因素和後天環境的影響時仍可能產生偏誤。第二種策略是使用堂兄弟姊妹或表兄弟姊妹的親屬資料。在控制相同的家族基因下，成長在不同家庭下的小孩成人後的表現即代表後天環境的影響(Goldberger, 1979; Behrman and Taubman, 1989)。然而，這個方法的缺點是這些小孩的父母，彼此是有血源關係的手足，小孩間相似之處不僅是遺傳基因而已，培育他們的家庭環境也可能近似，忽略這項因素則會高估先天因素帶來的影響。第三種策略是採用親生和領養子女的資料，這是晚近較為熱門的研究趨勢(Plug and Vijverberg, 2003; Bjorklund, Lindahl and Plug, 2006; Bjorklund, Jantti and Solon, 2007)。在領養小孩的過程符合隨機分派的假設之下，被領養子女的基因與領養父母無關，因此領養父母對養子或養女教育的跨代移轉效果即可詮釋為來自後天領養家庭的環境。本研究即是採用第三種實證策略進行分析。本計畫將使用合併台灣內政部「戶籍檔」、「出生檔」以及「大學聯考榜單與成績」所建立的兩套全國性大型資料庫，來探討父母對小孩教育的跨代移轉效果。這項資料的特色是除了可串聯出親生父母和親生子女之外，同時可串聯出被領養子女的領養父母和親生父母，是過去探討本項議題極為少見的珍貴資料庫。截至目前為止，國外研究中除了Bjorklund, Lindahl and Plug (2006, 2007)使用的瑞典資料稍具規模外，其餘皆是小樣本資料(以美國Wisconsin州一所高中畢業生的追蹤資料或韓裔美籍領養資料)的分析，亞洲國家的研究更是付之闕如。本研究將利用典型的跨代移轉模型(intergenerational mobility model)來探討先天因素和後天環境對教育的相對重要性。藉由領養資料中同時納入親生父母和領養父母教育程度進行估計，可以檢測究竟是「先天因素」或是「後天環境」(nature vs. nurture)對個人教育成就的影響較大。透過非線性模型的估計，我們將考量到先天因素和後天環境之間的互動性。針對文獻上提及的幾個可能造成跨代移轉效果估計偏誤的因素，包括領養不具有隨機性、小孩非出生後才被領養、被領養子女生父不詳的樣本選擇性，以及被領養子女是否會因為較得不到領養父母關愛，而使跨代移轉效果減弱等問題，本研究將一一採用相關估計方法來反覆驗證實證結果的頑強性。這是除了瑞典以外，文獻上首度以大規模的全國性資料進行的實證，本項計畫的研究發現將具有指標性的意義。 The relative importance of nature and nurture in intergenerational transmission of education is one of the most prominent research topics in the economics literature. Becker and Tomes (1986) and Haveman and Wolfe (1995) suggest that parental resources are linked to the educational attainment of children through human capital investment. Therefore, they put more emphasis on nurture in determining educational outcomes. Alternatively, Hernstein and Murray (1994) argue that it is nature rather than nurture that explains educational persistence across generations. Parents with more ability (measured as IQ) are highly educated, and the ability is transmitted genetically from parents to children. There are at least three possible strategies to shed light on the importance of the heritability of ability. The first strategy works with data on identical and fraternal twins. Since most of the twins share the same family environment. The differences between fraternal and identical twins identify the genetic contribution (Plomin, DeFies and McClearn, 1990). The main limitation of this approach is that most twin studies rely on small samples. Furthermore, nature and nurture estimates may be biased because identical twins are treated more similarly than fraternal twins. The second strategy considers relatives with similar genetic structure. The differences between relatives raised in different families are used to measure the environmental impact (Goldberger, 1979; Behrman and Taubman, 1989). The drawback of this approach is that children who are relatives share more than just genes (such as the similar household environment), which results nature estimates upward-biased. The third strategy, which is used more recently, compares children who are their parents’ own offspring to children who are adopted (Plug and Vijverberg, 2003; Bjorklund, Lindahl and Plug, 2006; Bjorklund, Jantti and Solon, 2007). We use unique Taiwanese data with information on adopted children’s biological and adoptive parents to estimate intergenerational mobility associations in education. We combine the household data, birth data as well as the college entrance examination data in Taiwan. To our best knowledge, there is no study on this subject based on such a nationally representative and comprehensive databank, except for the Swedish study by Bjorklund, Lindahl and Plug (2006, 2007). This study uses the prototypical model of intergenerational mobility to investigate the relative importance of nature and nurture in education. In addition to liner model, we also consider the nonlinear model by including the interacted effect between adoptive and biological parents. Several problems involved in using adoption data could lead to misinterpretations. They are (1) nonrandom placement of adoptees to their adoptive families, (2) not all adoptees are adopted as babies, (3) many adoptees have unknown birth fathers, and (4) adoptees and adoptive parents are different from other children and their parents. To ensure the robustness of our findings, we will adopt several strategies to examine the impact of each of these four problems. 研究期間：10008 ~ 10107 |
