家庭議題與小孩教育關係之三篇實證研究

Three Essays on Family Issues and Children's Education

陳妍蒨(Yen-Chien Chen)

指導教授 ： 劉錦添

國立臺灣大學/社會科學院/經濟學研究所/博士(2009年)

https://doi.org/10.6342/NTU.2009.00667

摘要

此論文包含三篇探討家庭議題對小孩上大學機率影響的實證文章：

第一章：The Impact of Sibling Sex Composition on Women's Educational Achievements: A Unique Natural Experiment by Twins Gender Shocks
在重男輕女的經濟社會中，兄弟的存在可能會排擠掉父母親對女孩教育的投資，尤其是在父母親有時間或金錢上的限制的情況下。然而，除了資源上的競爭，兄弟也可能會帶給女孩正面外部的效益。我們以兩套全國性的資料，出生檔與大學聯考檔，將兩個檔案進行合併，並選取第一胎為雙胞胎作為分析樣本，探討龍鳳胎（女-男雙胞胎）中女孩的教育程度是否因為雙胞胎弟弟（或哥哥）的存在而有所減少（與女-女雙胞胎來的女孩們相比較）。為了避免墮胎的行為會影響父母親選擇小孩的性別，我們將分析樣本限定在出生於1985年台灣優生保健法實施前，亦即墮胎合法化之前的雙胞胎。實證結果發現，雙胞胎兄弟的存在並不會影響雙胞胎女孩上大學的機率。這個結果也顯示兄弟的存在可能對女孩的教育有正面的外部效益。本文為第一篇以雙胞胎樣本估計手足性別對小孩教育程度影響之文章，且因本文所用的雙胞胎樣本為全國性母體資料，估計出來的結果將更為精確。

第二章：Separate Effects of Sibling Gender and Family Size on Educational Achievements – Method and First Evidence from Population Birth Registry
由於小孩性別及小孩數目同為影響父母親分配給每個小孩資源時的重要因素，所以文獻在探討家中兄弟姊妹性別對小孩教育的影響時，通常也會考慮家中小孩數目變數。然而，同時考慮這兩個變數所估計出來的結果，將會使性別的係數無法正確地衡量其對小孩教育的影響。原因是來自於，父母親決定生育的小孩數目時深受家中（較年長）小孩性別的影響。因此同時衡量這兩項變數對小孩教育的影響，將會使估計出來的性別效果包含：「直接」效果，以及「間接」效果。其中，間接效果是指性別對小孩教育的影響是間接透過對小孩數目的影響而得到，這個效果幾乎被所有的文獻忽略。本篇指出在這種情況之下，即使小孩數目沒有內生性的問題，也必須用一個工具變數來估計。此外，已有另一支文獻主張小孩數目有內生性問題，因此工具變數方法也是探討小孩數目與小孩教育之間因果關係的重要方法。綜而言之，以工具變數估計家中小孩數目的方法，不僅可以探討小孩數目與小孩教育之間的因果關係，也可以去除掉間接性別效果的問題。本文為第一篇以工具變數方法同時估計出有意義的性別效果與小孩數目效果之文章。我們以全國性的出生檔以及大學聯考檔，得到數量龐大的小孩樣本。並選取墮胎合法化（1985年）之前的第一胎單胞胎作為分析樣本，檢視第二胎小孩的性別以及家中小孩數目對其上大學機率的影響。文中所使用的小孩數目之工具變數是第二胎是否為雙胞胎的虛擬變數。我們實證結果發現，除了資料中最年長（1978年出生）的女性樣本之外，兄弟姊妹的性別以及小孩的數目對小孩上大學的機率無統計上顯著的影響。

第三章：The Impact of Unexpected Maternal Death on Education -- First Evidence from Three National Administrative Data Links
對一個學齡中的小孩而言，父母死亡無疑地是他（她）所能面對最傷痛的事情，且會對其未來身心發展造成長遠的影響。然而，由於資料的限制，以及死亡與一些無法觀察到的特性息息相關（例如父母死亡之前小孩就可感受到莫大的精神壓力），實證上難以探討失去父母及小孩教育表現這兩者之間的因果關係。本篇論文採用的資料是結合台灣全國性的出生檔、死亡檔以及大學聯考記錄檔，這三項全國性的資料可克服之前文獻所面臨的問題。我們以父母親的「意外死亡」來作為死亡的分析基準，並選取家庭中兄弟或姊妹樣本，以家庭的Fixed-effects模型來分析失去父母親對小孩上大學機率的影響。這個估計方法可以解決父母親死亡的內生性問題，因為在父母意外死亡之前並不會如同病死前一般對小孩有精神及壓力的影響；且Fixed-effects模型可以控制迴歸式中大部分無法觀察到的家庭因素。我們實證的結果發現失去母親對小孩上大學的機率有明顯負面的影響，但失去父親則在統計上並沒有無顯著的影響。這個實證結果與Behrman (1999)的發現大致相同。

關鍵字

小孩教育 ； 重男輕女 ； 性別選擇墮胎 ； 家庭內資源配置 ； 手足之性別效果 ； 手足總數效果 ； 父母意外死亡

並列摘要

This thesis includes three empirical studies on the effects of family issues on children’s college enrollment:

Chapter1: The Impact of Sibling Sex Composition on Women's Educational Achievements: A Unique Natural Experiment by Twins Gender Shocks
In a pro-male biased society, brothers may reduce the parental attention and investment received by female siblings, when parents face time or financial constraints. But brothers may also cause positive externalities. This paper tests whether women have fewer opportunities to attend college if they have a brother rather than a sister. We use matched administrative population data from a highly sex-imbalanced economy, Taiwan, and select twins at first birth as analysis subjects. To ensure our estimates are not confounded by sex- selective abortions, we exploit the fact that twin sex is purely random, given the sex of the other twin, once we limit the data to time periods in which abortion was illegal and technology was unavailable to abort one of the two twins. The estimates show that the birth of a male twin sibling, relative to a female, has almost no impact on women’s or men's college enrollments. If there is any effect, it only arises among females born in rural areas, or those born to mothers whose educational levels were a high school diploma or beyond, and even then it is small and statistically insignificant. This is the first paper which investigates the sibling gender effect by twinning shock. The estimates are precise because of the large number of observations in the population data. These results point to the importance of accounting for positive externalities created by a son's birth, in studies on sibling rivalry.

Chapter2: Separate Effects of Sibling Gender and Family Size on Educational Achievements – Method and First Evidence from Population Birth Registry
Because family size is endogenous and causally depends on the sex composition of early-born children, controlling for family size causes the estimated (total) sibling gender effect to include the direct effect, and the indirect effect mediated through family size. Therefore, previous evidence on sibling gender effect has not been well-identified. We show that separating the direct effect of sibling sex composition from the effects of family size requires an instrumental variable for family size, even if child gender and family size are both exogenous. By using a unique administrative data from Taiwan, we demonstrate how Instrumental- Variable Methods provide a way to resolve the problems of endogeneity and causal dependence of an important covariate (family size) on treatment status (the sex of the subject's subsequent sibling). For the purpose of this study, we construct a panel data of families using Birth Registry of the entire Taiwan and match it with College Entrance Tests records to obtain children's college attainment. This unique data provides accurate measures of educational outcomes, complete family size, and sibling sex composition. We minimize the incident of sex-selective abortion by focusing our analysis on cohorts who were born prior to abortion legalization and prior to widespread of technology for prenatal sex determination. We use the occurrence of twining at the second birth to instrument for family size, after controlling for birth weight residues. Our estimates show that neither sibling gender nor family size matter, except for firstborn girls who were born in the earliest year in our data, 1978, and the effects are small. After 1978, both effects are nearly zero.

Chapter 3: The Impact of Unexpected Maternal Death on Education -- First Evidence from Three National Administrative Data Links
The death of parents is one of the most traumatic events for school-aged children. It can have a long- lasting effect on a child's cognitive ability and socio-emotional development. However, because of data limitations and unobserved parental factors, including unobserved pre- existing health conditions prior to the actual death, the there is little existing evidence of any causal link between parental deaths and children's educational outcomes is little. Based on an unique data set, which is constructed by linking three national administrative datasets -- the Birth Registry, Death Registry and College Entrance Tests’ records from Taiwan--, we can exploit unexpected parental deaths in a family fixed-effects model in order to compare the difference in education between siblings who lost parents unexpectedly before versus after the age of 18. Our empirical strategies can remove the most of the problems caused by unobserved characteristics. We find that losing a mother has a strong impact on a child's possibility of entering college, while losing a father has no significant impact. Consistent with the suggestion in Behrman (1999), our evidence implies that a mother's role of in child rearing a child has a stronger impact on a child's outcome, than a father's role as bread winner.

並列關鍵字

Children’s education ； son preference ； sex selection abortion ； within-family resource allocation ； sibling rivalry effect (sibling gender effect) ； family size effect ； parental unexpected death

參考文獻

Mortality and Parental Deaths on Primary Schooling in North-Western Tanzania,”

Journal of Development Studies, 41(3):412-439.

Ananat, Elizabeth O. and Michaels, Guy. forthcoming “The Effect of Marital Breakup on the income and Poverty of Children?” Journal of Human Resources..

Angrist, J. D., Imbens, G. W., and Rubin, D. B. (1996), “Identication of Causal Effects Using Instrumental Variables,” Journal of the American Statistical Association, 91:444-472.

Angrist, Joshua D., William N. Evans. 1998. “Children and Their Parents' Labor Supply: Evidence from Exogenous Variation in Family Size.” American Economic Review 88(3): 450-477.