[EB] research

[Frank Aarebrot]

At 15:30 17.01.96 EST, you wrote:
>Hello Fellow Eurob Researchers:
>
>I am a professor at the univ of central florida who is utilizing the eurob
>studies in a study of the sources of attitudes toward economic reform.
>This is my first time employing the studies.  Can any of you who are more
>experienced w/ the studies alert me to any major problems w/ using the data
>sets?  I don't detect any at this early stage.
>Is there anything in particular I need to know about the studies?
>Thanks for your time.
>g hall
>
>
Hi, 


My name is Frank Aarebrot and I am an associate professor at the Department
of Comparative Politics of the University of Bergen, Norway. I have been
involveld in the European BIG (Belief in Governmont) project which used
Eurobarometers 37 through 42 for all European countries. I worked mainly
with data for the EFTA countries and the new democacies in eastern, central
and south eastern europe.

Whern getting into ths material, be prepared for most of the problems
associated with secondary data analysis: Unstandatdized coding og background
variables, varying sets of variables across contries. A a general rule the
attitudinal questien have the highest consistency across countries and are
thus of the highest quality. Unfortunately concrete variables relating to
political behavour such as Voting preference by party is NOT standard in the
EBs, and are thus available only for some countries, and in some cases fore
some barometers in one country, but not for others brometers for the same
country.

Sometimes the documentation is more extensive than the data. My advise is
threfore print out the variable labels and value labels from the SPSS files
and use them as codebboks.

In some cases several countries' files have been merged for the same EB.
These files will often contain massive strings of missing data for questions
asked in one country but not in another, or one variable for party
preferences in one country and another variable for party preference in
another country within the same merged file. Contrary, a variable such as
femily income may be fylly merged, containing almost no missing data,
irrespective of the fact that the countries in the file have different
currencies. My advice is to run freqency distribution for all variables you
intend to use and look for systematic patterns of missing cases and curious
"camel hump" distributions in variables where one would expect approximate
normal distribution. I chose to "unmerge" these merges files, creating a
system of one file for each EB for each country. It you need a Europe-wide
master merged file it is better to create it yourself after you have
familiarized yourself with the data so that you know which variables tend to
be available generally and consistenty coded. i.e. income can be
recalculated into e.g. US$ using the exchange rates and THEN merged.

Finally, in planning your research program. Estimate about 50 percent more
time for nitty-gritty data cleaning than you would for say a normal US
election study.

Good luck, you need it!


Frank Aarebrot