The detailed educational attainment variables used to construct the measures used in this analysis (Any Post-Secondary, University) differ slightly in the SLS and the YITS, since the YITS allows for a greater number of possible responses than the SLS. [...] The alternative set-up would have been to use a multinomial logit model, but the interpretation of the results would have been less straight-forward and the replicate weights available in the YITS rendered the implementation of such an approach problematic. [...] The rates thus rose everywhere, but least for the lowest parental education group, the most for the highest group, and in between these for the other groups. [...] Quebec, for example, had among the smallest tuition rate increases, and Nova Scotia among the greatest, but the increases in participation rates were relatively small in the former and high in the latter. [...] In any event, the results point to the need to look at participation rates—and the effects of tuition rates on participation rates—in a broader framework which takes these other factors into account.18 V. The Models V.1 Specification of the Models We employ a simple linear probability regression model, treating the two access measures—Any Post-Secondary and University—as the dependant variables.19