What is the best statistical test for measuring the magnitude, rather than the significance, of a sample mean?

I have an opinion survey (so most of the data is an ordinal level of measurement) that is based by geographic location—for example, ZIP code. We currently use the SPSS-recommended t-test to test for significance of change of the responses for respondents in the ZIP code from a survey performed three years ago to a survey just completed. What we are looking for is an appropriate statistical test to measure the strength or magnitude of that change, if any, in addition to finding out if it is statistically significant.