I have mentioned the different literary styles of party manifestos earlier. Today I simply wanted to share a gem I came across: “Bei Ausbildung, Jobchancen und Wohnungssuche begegnen viele mehrsprachige Menschen nach wie vor Barrieren.” (In English: When it comes to education, job opportunities and house hunting, many multilingual people are still met by barriers.)
Pause for a second. We’re looking at parts of the party manifesto about immigrants. To the authors words like foreigner or immigrant must have sounded too negative. They surely don’t mean the discrimination of polyglots because they speak many languages; they mean discrimination of foreigners who may not speak German (well).
This sentence is a gem as it illustrates the current debate on immigration: the tendency to see it in black and white picture. Immigrants are either seen as problems (I guess that’s black), or problems that come with immigration are denied (that’s white). Obviously there are many (more differentiated) positions in between, but somewhat they do tend to disappear in the public debate, and worse: among many academics.
My paper on the role of source language in the automatic coding of political texts (Wordscores, dictionary coding) is now available online. I make use of Swiss party manifestos to examine the impact of source language on party positions derived from the manifestos: does it matter if a French or German manifesto is used? The conclusion is that both stemming and particularly stop words are important to obtain comparable results for Wordscores, while the keyword-based dictionary approach is not affected by language differences. Replication material is available on my Dataverse.
Edina Szöcsik and Christina Zuber have conducted an expert survey on party positions on ethnonationalism (full data are available).
I wondered whether the positions captured in the EPAC were so different from what other expert surveys cover. This figure is a simple scatterplot matrix comparing dimensions in EPAC with two dimensions in CHES (Chapel Hill). I only include the dimensions that correlate highly, but what is striking is that such high correlations can be observed.
The dimensions are: ethnonationalism (“ethno” from EPAC), cultural autonomy (“cul” from EPAC), territorial autonomy (“ter” from EPAC), rights for ethnic minorities (from CHES), and political decentralization to regions/localities (from CHES).
I did a quick factor analysis, and it seems that these five dimensions are sufficiently summarized in a single factor (p=0.0129; compared to p=0.159 for two factors).
In this working paper (The Paradox of Manifestos) Ian Budge replies to some of the methodological critiques of the MARPOR/CMP/MRG manifesto projects. The paradox lies in the contrast between the positive research experience of most users of the manifesto data, and the at times rather harsh methodological critiques of manifesto data.
Unfortunately the paper is quite selective in which critiques it engages with (table 2 is rather short). The biggest issues I have come across with the data in question are a rigid coding scheme (this of course has advantages, but the data can struggle to reflect the situation on the ground adequately), and party rankings that defy common sense and expert judgements. In my view the many happy users Budge identifies are a sign of good enough data (not necessarily good data), and also of the extensive coverage of the data.
Having tried and compared different methods to measure party positions, I have serious doubts whether we’re even close to measuring party positions with precision — or do precise party positions exist at all? No, I don’t want to give up on measuring party positions, after all the different methods correlate enough to agree on the ranking of parties. We should, however, always express the precision and error in measuring party positions, not just the point estimates.
Two weeks ago I wrote about different methods to obtain party positions on immigration. I compared different methods, including manifesto-based ones. You can check out the figures, including two not included in the research note.
With the short extracts on immigration, I was unable to get realistic party positions on immigration using Wordfish. Interestingly, using just a few additional documents (the way CMP controversially does it) seems to lead to realistic estimates.
Part of the research note is about a retrospective survey to place parties. I basically asked my colleagues how the parties are placed today and in the past 20 years. None of them works on parties. Taking the average does yield surprisingly good estimates, as outlined in the research note. Unsurprisingly, the standard deviations are large, but they are not uniform or simply larger the further back we go.