I'm looking for information on comparing statistics, such as from a survey, from different years. The population, for example "all current members of the YMCA" might change a bit from year to year due to new and dropped memberships--does this have to be taken into account? Also, the sample size of each year's survey might be different due to different response rates.

The kinds of comparison I might want to do would involve comparing categorical variables from year to year, such as, "What percentage of men members responded that they smoked cigarettes in the 2007 survey compared to the 2006 survey." Given that the sample sizes may be different, and that the population may be slightly different due to new and dropped memberships (does that matter?) what tests would I use to figure out what constitutes a statistically significant difference in percentage between two different years? How would I figure out if an 8% difference between two years is statistically significant?

I've done research into this question, but most materials I've come across are comparing statistics from one sample or samples taken close in time. I came across a mention of using two-way tables for dependent samples--I assume these samples may be dependent because I'm looking at roughly the same population from year to year, and comparing the same segments of the population (i.e. men who smoke, women who smoke, etc)--would the two-way tables for dependent samples be a test I would want to use, are there any other tests and considerations to keep in mind? Thanks for any help.