Commuting or re-location costs could be an in important influence on students' university choices and might even deter some from going to university. The barriers presented by these costs may be high for lower-income students, and students for whom there are cultural incentives to remain in or close to the parental home. If this is the case, then the geographical accessibility of universities has an important bearing on differences in higher education choices for different income and ethnic groups, and, in turn, on their earnings and life chances. Existing evidence has shown that university places are not evenly spatially distributed in Britain. Research has also found that "non-traditional" students--those from backgrounds in which higher-education participation is emerging--cite the location of institutions as a factor affecting their decision to go in to higher education. However, it is easy to make the mistake of attributing behaviour to ethnicity, gender or income when these behaviours are really due to other differences, like academic achievement, or home location which will have strong bearing on if and where students go to university. In fact, there is no large scale, systematic evidence for the UK that shows that proximity to a university really matters for higher education participation or choice amongst universities, or that it matters more for specific ethnic or income groups. The authors' research looks at these questions using administrative data on the population of school leavers and university entrants in England. These data allow the authors to link the choices of students from different ethnic and income backgrounds to distances between home and university, whilst accounting for schooling, neighbourhood and other background characteristics. Their key findings are: (1) Universities are not evenly distributed around the country but 90% of locations have three institutions and 4000 first degree places within 100km; (2) Non-white ethnic groups and low-income students actually live closer to their nearest three higher education institutions and closer to their nearest three high-quality research institutions than their white and high-income counterparts. These facts suggest that disparities in geographical access are unlikely to be a source of disadvantage to ethnic minorities and poor students; (3) Home-to-university distance has only a tiny influence on the probability of participation in higher education, relative to achievement and other background factors. Their statistical models imply that doubling the distance to the nearest institution would reduce the probability of white female participation by at most 4.5% in relative terms--reducing the probability of participation at the mean from 28.4% to 27.1%. For males, the effect is only half that, but there are no systematic differences by ethnic or income group; (4) In contrast, distance is the strongest factor influencing university choice amongst those who participate. The probability that a student attends a specific university decreases by 8%-15% with each 10% increase in home-to-university distance. This distance cost is observed for all ethnic and income groups, but is highest for Pakistani and Bangladeshi girls and low income students, and lowest for Black students and those from Professional backgrounds; and (5) The influence of distance on choice of institution could make a difference to the type of higher education received by different demographic groups. This is a moot point for ethnic minorities, who have high participation rates at "elite" research intensive universities relative to whites, but provides a potential explanation for lower participation rates amongst women and low income groups in top ranked research universities. The findings therefore offer no support for the idea that improving the accessibility of higher education institutions is an effective route to raising participation. However, targeting the accessibility of higher-quality institutions could increase uptake of high quality HE places amongst suitably qualified students from lower-occupational status backgrounds. Such policies might include action to reduce the role of distance (distance learning) but also policies to encourage higher status institutions to undertake outreach activities further afield. In any case, the authors find no evidence to suggest that such a policy need be gender or ethnically targeted. One further important spatial implication from this work is that the type and quality of higher education in which students enroll is in part governed by the type and quality of local institutions, which in turn partly determines the skill composition of the local population. Given this, the local mix of institution types and quality could have a strong bearing on the quality and composition of the local human capital stock. (Contains 11 tables, 4 figures and 23 footnotes.)