V-Party data doesn& #39;t measure what parties do, or what they say they& #39;ll do. It measures what political scientists *think* of each party.

665 political scientists were polled in Jan 2020 to produce this graph. Its designed to reflect the ideology/expertise of mainstream polsci (1/ https://twitter.com/BrunoLeipold/status/1320662618193960962">https://twitter.com/BrunoLeip...

Big challenge with this is that the 665 are asked about specific countries & tend to be specialists in that country. V-Dem use various techniques to try to correct for this (ask the experts to answer questions for other countries too, & use Bayesian Item Response modeling) 2/

Clearly it has produced some pretty weird results here (Labour& #39;s manifesto being further left than MAS...).

But other problem is the use of heavily ideological terms as if they were neutral terms of measurement (exacerbated by Guardian changing & #39;liberal& #39; to & #39;democratic& #39;) 3/

Here& #39;s how they measure left-right. Note that & #39;active& #39; here ignores all the activity governments do to make/enforce markets. Also, use of more/less means implicit comparison within each country (more regulation in Britain would still leave us way less regulated than Germany) 4/

The illiberalism index then describes itself as being about "commitment to democratic norms", but two of the four components are actually about tone: "personal attacks on opponents" (check the image), and "explicitly discouraging the use of violence". (5/)

You can read the codebook here: https://www.v-dem.net/media/filer_public/04/d1/04d109be-3931-4a99-8ada-561c7c9b6ea0/v-party_codebook.pdf">https://www.v-dem.net/media/fil...

This was obviously put together with more survey expertise than I have, but the central challenge remains: if you poll people from a narrow professional group, you& #39;ll get results which reflect the ideology of that group.