We are doing a project on Race, Gender and Employment. We wanted to track race, gender, occupation, wages, benefits over time (1930-2012). I have heard that coding in the census changed dramatically during 1980 and after. Is this true and are there other items to be aware of when dealing with historical data?
It is true that coding structures have changed drastically over the time period you are interested in. However, the RACE variable is an excellent example of how IPUMS-USA integrates multiple coding structures to create a harmonized variable that is comparable across time. The primary issues that you should be aware of when comparing across large spans of time are the changes in question Universe that may make variables incomparable over time. For instance, the OCC1950 variable underwent multiple age limitation changes over the years. This means, that to produce a true comparison over time, you would have to restrict all of the data to the most restrictive universe. You will also want to study the sample designs of each sample you are using to make sure the samples themselves are comparable. For example, the 2000-2005 ACS samples do not include persons in Group Quarters.
I hope this helps.