During the dictatorship in Chile, the authorities sought to discourage political participation and implemented voluntary and cumbersome voter registration procedures. Together with compulsory voting rules already in place, this created the unique combination of voluntary registration and compulsory voting.
I learned this from a newly published paper (gated) by my friend Paulo Cox and his coauthor Alejandro Corvalan. If one guessed, e.g. on the basis of rational voter theory, that this rule would cause registration and turnout to plummet, one would be right. There is more to the story in Chile, though, as voters registered en masse to vote Pinochet out of power in 1988. After that, however, the prediction has it right, with the result of a gap in turnout of around 60% (ca 90-30) between older and younger voters in the 2009 election.
Corvalan and Cox’ main concern is the representativeness of the electorate, and they show that income is highly correlated with registering and voting among the young (when controlling for education, etc). As the young become old, this “class bias” might come to hold for all voters.
One person, one vote is a cherished principle, however, in most countries the representatives to the national legislative body come from certain fixed geographic regions. E.g., in the US Senators are elected from the states, while the members of the House of Representatives come from a congressional district.
In Norway, each of the 169 representatives in parliament, the Storting, comes from one of 19 counties. At the heart of the distribution of seats is a weighted sum of the number of inhabitants of a county and its area, calculated as follows: #inhabitants+1.8 x area in sq.m. So 1 sq.m. counts as 1.8 person. This is a systematic overrepresentation of people in counties with relatively low population density and corresponding underrepresentation of people in more densely populated counties.
How does this play out in practice? For the elections in 2005 and 2009, the seat allocation was based on the inhabitant count as of 2004-01-01. Below I have plotted the number of representatives per inhabitant against the number of inhabitants for all the 19 counties. Counties with few inhabitants tend to have more representatives per inhabitants. The most Northern county, Finnmark, is the outlier, with more than double the number of seats per inhabitant as several other counties. (Or in other words: Fewer than half the number of people behind a representative.)
Due to migration within the country and a small reform, the allocation for the elections of 2013 and 2017 is shifted somewhat. More populous counties will now be a little less underrepresented:
This topic is currently on my research agenda, so I will return to it in not too long. A seminal paper (pdf) by Samuels and Snyder, more countries on Andrew Gelman’s blog, discussions of representation in the US Senate on CrookedTimber and in the New York Times.
Update: for Norwegian readers: Discussions from the election in 2009, from Minerva, Aftenposten and Ferskvannsmandatet.