中文摘要：

评估轨道交通绩效的一个重要方面是考察轨道交通可达性的提高能否促进居民交通行为转变,尤其是减轻对小汽车的依赖。目前的研究结论不尽一致,且研究中存在的自选择问题尚没有引起足够的重视。基于2013年上海市80后群体的微观调查数据,分析轨道交通可达性对城市居民交通方式选择（小汽车购买和使用以及轨道交通使用）的影响,并利用样本的异质性来克服轨道交通可达性与交通行为之间的自选择。研究发现：居民居住区周边轨道交通可达性的提高,确实能够显著地降低家庭小汽车的拥有概率,并提高乘坐地铁通勤的可能性;但不能找到轨道交通可达性的提高会减少小汽车使用的证据。

英文摘要：

Increasing vehicular travel demand in urban areas has resulted in negative consequences such as traffic congestion and air pollution. It is widely acknowledged that the growth of auto ownership is a major cause for the deterioration of urban traffic conditions. In developing countries, auto ownership levels have grown substantially. In the mainland of China, the period between 2010 and 2014 witnessed an increase of private cars from 78 million to 145 million at the national level. In the meantime, many cities in the world, especially those in fast-growing developing countries, have made massive investments in rail systems and considered them an alternative to reduce individuals＇ reliance on automobile and mitigate traffic congestion and environmental impacts of transport. However, no consistent result has been concluded from related researches, and few studies have examined the relationships between transit access and auto ownership in developing countries. More importantly, available studies are far from satisfactory for grossly neglecting the self-selection problem. Based on a household survey of Shanghai residents who were born from 1980 to 1989, this paper analyzes the influence of rail transit accessibility on travel mode choice. Using Probit and Heckman two- step estimations, we find that, after controlling for demographics and other urban built environment variables, the improvement of rail transit accessibility can indeed significantly reduce the possibility of car ownership, and encourage the use of Metro. Besides, no evidence has been found that the improvement of rail transit accessibility can affect the use of cars. However, these findings may still suffer from residential self-selection, specifically those who prefer not owning vehicles may choose to live in transit-accessible neighborhoods. In order to confirm our findings, we adopt two approaches to solve this problem. The main idea behind the two approaches is to reserve samples whose residential locations are not totally the outcomes of their o