whereas over two-thirds of those with no qualifications did so.This impact of education on vote choices is also highly significant in themultiple logistic regression models. As log odds are not straightforward tointerpret, the last column in Table 2 shows the marginal effect of one standarddeviation change in each of the explanatory variables on the probability ofvoting Leave. This shows the strongest effect for education and age. Goingfrom A-level education to an undergraduate degree reduces the probabilityof voting Leave by about 10 percentage points, all other things beingequal. Similarly, a 50 year old is 10 percentage points more likely to supportBrexit compared to a 33 year old voter. Men are slightly more likely to voteLeave (2 percentage points), as are those with lower incomes and thosewho feel that their financial situation has deteriorated. These are very substantial differences, especially when it comes to the generation and educationgaps; however, the overall model fit is modest (pseudo R-squared of 0.09)