本文以台灣日治時期水道建設為例，連結當時總督府統計與現代人口普查及中老年身心調查資料，研究出生時期乾淨用水取得機會的增加，是否與日後的教育、健康與婚姻狀況有正向關係。我們發現，在OLS估計下，出生時水栓供水戶數（每千人）對日後教育年數、完成某一階段學業（如國中畢業）、仍維持良好婚姻與降低老年患病的機率等變數皆有顯著的正向關係。為了解決OLS內生性偏誤的問題，我們也提出以前一年的營業稅與雜種稅作為水栓供水戶數（每千人）的工具變數的作法。兩階段迴歸估計顯示，水栓供水戶數（每千人）仍與教育、婚姻與健康有顯著的正向關係，且2SLS的估計值大於OLS估計結果。整體來說，本文結果顯示嬰幼兒時期乾淨用水的取得確有長期影響，亦對早期嬰幼兒干預政策的成本效益分析，提供了討論的基礎。

We use Taiwan's pipe water construction in the early 20th century as a natural experiment. By linking hisrorical Japanese colonial era data and contemporary population census and survey of the elderly data, we examine whether the number of households with hydrants per thousand people at birth will affect education, health, and marriage status years later. OLS estimates show that the number of households with hydrants per thousand people is positively correlated with education, marriage status and the health of the elderly. To solve the endogeneity problem, we use one-year lagged sales and miscellaneous taxes as instrumental variables. 2SLS still shows the positive correlation between clean water and later developmental outcomes. And its effect is larger than that of OLS. Overall, our evidence confirms the existence of long term effect of clean water access infancy, and provides a basis for the discussion of early childhood intervention policy.