A lot of people think that party polarization could be mitigated if states reformed their system of nominating candidates. This is based on the assumption that those who vote in primaries are more ideologically extreme than general-election-only voters. The extreme primary voters, according to this theory, nominate extreme candidates, which leaves the moderate general-election-only voters with no moderate candidates to vote for. This results in the election of candidates who are more extreme than the majority of voters actually prefer.
John Sides points out that the political science consensus suggests otherwise. Among other things, he points to the chart above, which is based on a large survey of voters in the 2008 election. The data demonstrate that there is in fact little ideological difference between primary and general-election-only voters. For the most part, primary voters are simply more interested in politics than are general-election-only voters:
Those who voted in the primary were clearly more interested in politics but did not have very different views on issues (with the possible exception, for Republicans, of raising taxes on the wealthy). Other research finds a similar pattern; see here or here or here. Given these findings, increasing primary turnout would not necessarily create a very different electorate and therefore different incentives for candidates or incumbents. [read the entire post here]