國中偏差行為對未來人際關係的影響

菸酒行為不只會損害身體健康也會造成其他的負面效果，如失業、人際關係惡劣等，但是依然有許多人不願意戒除這些行為，甚至在台灣未成年人抽菸或喝酒的比例逐年大幅上升，並有研究指出偏差行為常發生於國、高中職階段。雖然過去研究多有提出家庭、學校、同儕效果等人際關係對於抽菸與喝酒行為的影響，但鮮少有研究指出這些行為是否又會對於同儕或周遭的人際關係帶來影響，而且這些研究大多以短期影響為主，並不知道國中時期的偏差行為是否會帶來長遠的影響。　　本文利用「臺灣青少年成長歷程研究」(Taiwan Youth Project, TYP)的長期追蹤資料，探討台灣學生在國中時期的飲酒、抽菸行為，是否對未來高中時期、大學時期、成人階段的人際關係造成影響，並將人際關係加以細分愛情、友情及親情等三方面。然而飲酒、抽菸行為與人際關係存在互為因果的內生性問題，使估計出的係數可能存在偏誤，因此本文另外透過兩階段工具變數法解決內生性造成的偏誤。　　工具變數法的第一階段估計中，採用受訪者滿不滿意與國中導師之間的關係 (IV1)、導師喜不喜歡該受訪者 (IV2)並同時考量兩個變數，其F統計量皆大於10，具有顯著的解釋力。研究結果發現在第二階段估計，與未考慮內生性的最小平方法估計 (OLS)相比，可以得出OLS的係數結果會因為內生性而低估。　　解決內生性後，國中偏差行為對高中的友情、親情皆為負面影響，唯在愛情上容易交到男女朋友的正向顯著。在大學時期國中偏差行為對大學時期的友情影響不大，但在親情上有持續的負面效果，受訪者與家人關係相對不好。大學愛情主要為負面影響，國中偏差行為容易導致未來大學時期會相對容易分手、失戀。最後長期的成人階段，友情、親情與愛情皆顯示出負面顯著效果，代表國中時期的偏差行為依舊會影響到多年後，與人相處的狀況與好壞。

關鍵字

偏差行為；人際關係；內生性；工具變數法；台灣青少年計畫

並列摘要

Smoking and drinking not only damage physical health but also cause other negative effects, such as unemployment, poor interpersonal relationships, etc. However, there are still many people who are unwilling to quit. Moreever, the smoking rate or drinking rate in Taiwan has increased significantly in recent years. Some studies have pointed out that risky behavior often occurs during junior and high school. Although many previous studies have examined the effects of family, school and peer effects on smoking and drinking behaviors, only few studies have investigated whether these behaviors affect interpersonal relationships in the short-run. We don’t know whether the risky behaviors have long-run effects on interpersonal relatiopnship. This research uses the longitudinal data of the Taiwan Youth Project (TYP) to explore whether the drinking and smoking behaviors of Taiwanese students during their middle school years have an impact on their future interpersonal relationships in the long run. We consider three types of interpersonal relationships: love, friendship and family. To consider the endogeneity of risky behaviors and interrelationship, this paper uses a two-stage least squares regression to solve the endogeniety bias. This paper consider the relationship between the interviewe’s dissatisfaction and the tutor in middle school (IV1) and the tutor’s liking for the interviewee (IV2) as the instrumental variables of risky behaviors. The F statistics are all greater than 10, which have significant explanatory power. After solving the endogeneity, the risky behaviors have a negative impact on the friendship and family in high school but positive effects on the probability of having a girl friend. In the college period, the risky behaviors have little effects on friendship, but have negative effects on family relationship and love relationship. In the adulthood, risky behaviors have negative impacts on friendship, family relationship and love relationship. The IV estimates indicate that risky behaviors in middle school have significant impacts on interpersonal relationship in the long run.