There are no simple answers to your  question since you did not post your full logistic regression. Here are several things to check:

1)Have you run a univariate analysis before the logistic regression
2)how many variables are finally in your model
3)how about the missing values of each variable
4)any multicolliniarity problems
5)any possible interactions between the variables?

麻烦您给看看谢谢

Just based on your results, it seems nothing specially wrong. You could have a quick check of the interactions between education and umemployment, or between education and city, or between city and unemployment. If there are interactions, that may be the reasons. Otherwise you have to accept the results for your population.