
129Testing the Parallel Regression Assumption
nonparallel model. Education has a signicant, positive association with the probability of
mixed support for welfare spending in the nonparallel model. For both age and education,
relaxing the parallel regression assumption leads to different substantive conclusions.
Based on the parallel model, we would conclude that age and education are statistically
unrelated to welfare attitudes. However, based on the nonparallel model, we would con-
clude that education and age have a signicant, positive association with the probability of
mixed support for welfare spending.
We examine the implications of relax ...