This is an annotation to my previous post.
I don’t want to get to technical in these blog posts, so I’ll try to keep the methodological description of the analysis at a minimum. Despite saying this, I still think it is necessary to give a little insight into what i did in the analysis.
The software I used is Stata.
Here you can find access to the datasets from WVS and EVS. Both are free of charge and can be downloaded as separate cross-sectional datasets from the different waves or as time-series files for different statistical analysis software.
In my analysis I combined EVS data for the 28 current EU countries from 4 waves between 1981-2010 and WVS data for the US from 6 waves between 1981-2014. The time between waves and the date of the different waves differ between EVS and WVS. This had to be kept in mind when analyzing the data.
The WVS also includes some EU countries. I excluded this data from the analysis to keep the EU data as contingent as possible. Including the data from a few select EU countries – but not all – would have made the comparison between means for EU and US less precise rather than adding to the quality of the analysis (at least as long as I assume, that the EU countries differ from one another, which I do).
I included every country that is an EU member state today, regardless if they have been a member at the time of the survey. I did this because I wanted to achieve a complete picture of the change in values in europe – or at least the EU-members – over the last decades. This decision obviously can be critized and I’m still split on the issue myself. If and/or how the membership in the EU – or the candidate status – changes values and beliefs in a country is another interesting question for a different time.
What I did with all questions is to recode the results into a dummy-variable that shows if the respondents rather agree with the statement or if they do not. I made the decision to treat the answer “Neither” as not agreeing with the statement. After this, I aggregated the means on a country level and the country means on a EU level. I then compared the results in a graphical analysis.
If anyone is interested in the specifics of the analysis, feel free to ask in the comments.
You can find the annotated Stata do-files here:
Preparation and merging of the datasets: https://drive.google.com/file/d/0B_zYsItgIBT8bkwzcUh4NTRzbG8/view?usp=sharing
Another problem that emerged quickly was, that the 3 questions were not included in all waves, especially in the US. So the earliest data for the EU begins in 1993, for the US in 1998. Also question 2 dropped out of the US survey after 2004 and question 3 was included as late as 2014. This limited the analysis somewhat, but I tried to work with what the data could provide.