In the run up to this month’s parliamentary elections, political parties and candidates answered 43 questions on various policy issues. On the site, voters take the exact same questionnaire. The site then gives users “a ranked list of candidates & parties that best match their political profile.”
The software flags parties and candidates in the voter’s electoral district whose views align two-thirds of the time using “geometric matching”. The matching score does not indicate that the user is in “absolutely in agreement with the answers of a candidate”, according to the site’s FAQ. So, for example, “a match of 70% does not mean that a candidate has given the exact same answers as the user for 70% of the questions.”
Some of the 43 questions asked are:
“Should the number of working hours per week be lowered without reducing wages?”
“Are you in favour of forcing municipalities to tax empty apartments?”
“Are you in favour of lowering the age of retirement?”
“Are you in favour of extending the offer of international programmes within public schools?”
“Are you in favour of the automatic entry of foreign inhabitants into the electoral registers for local elections?”
“Should a public television broadcaster be introduced?”
“Should a single electoral district be introduced for the general elections?”
Your correspondent was slightly surprised by his suggested matches. He did not expect his top matched candidates and party.
More surprising were how closely the scores fell. Smartwielen highlighted 21 candidates from three parties with matching scores between 66.5% and 72%. It listed three party lists with matching scores that fell within 7 percentage points of each other; six party lists had matching scores within 12 percentage points of each other (between 57.3% and 68.9%).
Part of this probably reflects a certain level of consensus across Luxembourg parties (and quite possibly your correspondent holds relatively bland political opinions). But it also reflects the limitations of voting advice applications, and the fact that not all candidates have taken the questionnaire (62% had as of 19 September).
The political scientists behind this app pointed out that the site serves as research resource and does not aim to provide a final voting recommendation. “The goal of Smartwielen is not to encourage users to vote in a certain way, but rather to invite them to seek out more information,” the site’s FAQ stated.
Under that lens, the site is an excellent starting point for voters.
The app is used anonymously, although the data is collected by the university “for research purposes”.
As of this writing, Smartwielen had provided 93,811 recommendations to voters.