應用因果推論評估社會福利政策之效果：三篇論文

Abstract

本研究進行了三篇實證研究，運用因果推論與行政資料評估台灣三項社會福利政策之效果，分別為老年年金法定請領年齡、育嬰留職停薪津貼以及生育津貼。 第一篇研究探討勞保老年年金的法定請領年齡延後對退休金請領和退休行為的影響。2009年，台灣政府宣布法定退休金請領年齡逐步從60歲提高至65歲，這一政策影響了1958年1月1日及以後出生的勞工。對於1958年出生的一代而言，領取全額退休金的年齡提高至61歲，而領取減額退休金的年齡則提高至56歲。我們採用斷點回歸方法比較1957年與1958年出生者在請領退休金和退休行。研究結果顯示，提高法定請領年齡一年使得領取退休金的時間延遲0.55年，退休年齡延後0.38年，同時，61歲及以後才退出勞動市場的比例也增加了16.6%。收入和流動資產較低，或儲蓄較不積極的個人，延後退休與請領退休金的時間較長；此外，規模較小或聘僱年長員工比例較高的企業，其員工請領退休金和延後退休的幅度也較大。 第二篇研究分析帶薪育嬰假津貼對生育決策的影響。帶薪育嬰津貼的金額為育嬰假前六個月的平均勞保薪資的60%，且由於投保級距的存在，津貼亦受到上限限制。換言之，對於最高投保薪資的受益者，其津貼與薪資比例會隨薪資增加而下降。我們利用此津貼的外生變異來評估政策效果，使用的方法為拗折點回歸。實證結果顯示，津貼的所得替代率每增加1%，女性在當年或次年生小孩的機率就會提高1.06%，尤其是在女性員工比例較高的公司工作的女性，其影響更為顯著。然而，男性的生育行為並未受到育嬰津貼增加的影響。 第三篇研究旨在探討生育津貼對生育行為的影響。為了刺激生育，台灣各縣市政府對新生兒家庭發放一次性現金補助。2015年時，部分縣市的生育補助顯著增加，而其他縣市則維持原有水平或僅微幅調整。我們透過此生育津貼的外生變化，以及差異中的差異法來評估生育津貼對生育行為的因果關係。研究結果顯示，生育津貼提高對生育的影響相當有限：每增加10%的生育津貼僅使無子女婦女的生育率增加0.9%，對已有子女的婦女則沒有顯著影響。此外，縣市生育津貼金額的差異會導致跨縣市的人口遷徙，這一結果表明，在評估具有地域差異的生育津貼對生育行為的影響時，必須考慮到國內遷徙，以避免估計誤差。

This dissertation encompasses three studies that utilize causal inference methods and administrative data to analyze the impacts of welfare policies. The first study examines the effects of increasing the pension eligibility age on retirement decisions and benefit claiming. The second investigates the causal effects of paid parental leave benefits on fertility behavior. The third study evaluates the influence of baby bonus programs on childbearing patterns. Chapter 1 analyzes the impact of increasing the pension eligibility age on benefit claiming and retirement behavior in Taiwan. In 2009, the Taiwanese government raised the pension age by one year: from 60 to 61 for full benefits and from 55 to 56 for early reduced benefits. This policy change variably affected individuals based on their birth dates, particularly impacting those born after January 1st, 1958. Using a regression discontinuity design, we compared individuals born at the end of 1957 with those at the start of 1958. Our findings reveal that raising the pension eligibility age led to a delay in benefit claiming and retirement, with a 0.55-year delay in claiming and a 0.38-year postponement in retirement. The employment rate post-61 also increased by 16.6% for the affected cohort. Further analysis showed that lower-income/liquid asset individuals and passive savers are more likely to delay retirement and claiming benefits due to increased eligibility age. Additionally, employees at smaller firms or those with more senior staff exhibited greater delays in claiming and retirement. Chapter 2 investigates the causal effects of paid parental leave benefits on childbearing decisions in Taiwan. The parental leave policy provides benefits calculated as 60% of average insured salaries from the six months preceding the leave, with a maximum threshold. Employing a regression kink design to estimate the impacts of discontinuous changes in the benefit schedule on childbirth, our findings reveal a 1% increase in the replacement rate raises women's probability of giving birth in the current or following year by 1.06%. Furthermore, increased benefits largely increase the chance of giving birth for women working in female-dominant and larger firms. In contrast, the influence on men's fertility remains negligible. Chapter 3 examines the impact of bonus expansions on childbearing behavior. Local governments in Taiwan have implemented spatially varying one-time cash transfers to families with newborns as a measure to encourage fertility. In 2015, some localities introduced significant increases in baby bonus amounts, with the average bonus rising by 15,000 TWD (approximately 500 USD). Employing a difference-in-differences design, we compare fertility patterns between women residing in counties that increased bonus amounts and those that did not. Our findings suggest the effect fertility is negligible: A 10% increase in bonus amounts leads to just a 0.9% increase in birth rates for childless women, with no significant effects for women who already had children. Furthermore, we find evidence that varying bonus amounts across counties incentivize domestic migration. Our results suggest that when evaluating the effects of financial incentives that vary across districts on fertility, accounting for domestic migration is crucial to avoid biased estimates.