Here's
a scenario with which you might be familiar. You're working on a
paper, and you know that the trend of what Americans think about some
aspect of your topic is likely to be helpful in making your point. It
could be anything from views on the legalization of marijuana, to the
importance of the separation of church and state, to the legalization
of abortion, to which candidate various groups will support for
national office. There seems to be no limit when it comes to opinion
polling. How do you get the data you need for your paper? What are the
options? Generally, people using public opinion surveys look for one of three things:
- An individual question that appears in one survey or that appears in several surveys over time (enabling a trend or time-series study). The sought-for elements are the question text (with potential or actual responses) and the frequencies for each response. This aggregate data is also called "marginals" or "toplines." This is the "Yes 40%, No 30%, Don't know 15%, Braindead 5%, Other 10%" information.
- All the questions appearing in one survey. The safest way to retrieve a questionnaire's questions is to start from a known question in, say, iPOLL*, and search on that question's survey organization plus beginning and ending dates. iPOLL*, the Odum Institute* question database, and the Gallup Brain* offer a shortcut that might omit some of those modular questions recycled among concurrent surveys -- for each question retrieved, the database presents a hyperlink that assembles the questions.
- The raw data, or individual respondents' responses to each question, also known as microdata [where marginals are aggregate data]. Seldom if
ever would you want microdata, even if you were to say, "I want the
data for these questions." In most cases, what you really want are
marginals, or frequency of response. Generally, you will need microdata
only when you ask as a follow-up question: "I need to know how many
Hawaiians hated Ronald Reagan" or "Do poor, uneducated city dwellers
approve of school vouchers?" The trigger in these cases is the detail,
the subgroups. Other potential microdata uses would involve relations
among more than one question: "How many people who like mustard on
their pretzels would pay $500 for an Eagles ticket?" Generally, unless
you have viewed the survey's questions or codebook, it's likely not worth your
time, or the time of the person who's going to help you with using SPSS
or SAS, to re-process the raw data.
iPOLL (Roper Center), Penn Library Web
Odum Institute Public Opinion Poll Question Database, Penn Library Web
Gallup Brain, searchable via Sharon Black, Annenberg School Library. Gallup charges a pretty penny for access to their full archive question database. You will not get access by going through http://www.gallup.com/ either. As all the old Gallup polls and the main series of recent Gallup surveys appear in iPOLL, you can often get what you need without accessing the Gallup Brain.
There are other places to go for polling data. If you want to explore on your own, feel free to browse the University Library's Research Guide on Public Opinion Polls
If you have questions about this kind of research, please feel free to contact me.
