An interesting paper is Cardif and Klein (2005) which finds a ratio of 5:1 across tenure track staff in Californian universities in 2004. For comparison, the equivalent ratio across all Californian voters in the same year was 1.24%, and 1.06:1 across the US.
Some interesting points:
1) As of 2015, the Democrat to Republican ration among registered voters in California had lifted to 1.54%. Across the US, it has increased to 1.23:1 That is, the drift has been towards the Democrats. This rules out the possibility that the high proportion of registered Democrats is an artifact of a more tenure track university staff maintaining their views while the general electorate drifted to the right, if the high D:R ratio was not sufficient to rule that out in any event.
2) The D:R ratio varies substantially by faculty and by academic division, being highest (for the later) in the Humanities at 10:1 (see table 3), and lowest for Military/Sports at 0.7:1. The range of ratios can at least in part be explained by self selection to areas of interest based on political affiliation. Thus it is no surprise that Business has a low ratio at 1.3:1 while Social Sciences a high one at 6.8:1. However, except for the relatively small faculties included in Military/Sports, no academic division has a D:R ratio less than the general voting population of California, which puts paid to my theory that the disproportionate number of left leaning academics in social sciences was due to self selection.
3) Hard Sciences/Mathematics has a D:R ratio of 6:1, while Engineering has a D:R of 2.5:1. I consider these cases significant given that the subjects have no innate political content. The high D:R ratio in both relative to the general population, and in Hard Sciences/Mathematics relative to the academic population puts paid to any theory that citation bias or other academic biases is the cause of the high D:R ratio. It also makes very questionable the idea that the high D:R ratio is due to selection bias in recruitment.
That leaves several possible explanations. Of those the most favourable to conservative feelings is some variant of "those who can't do, teach". Put alternatively, the bias may arise from Democrat leaning people having a greater interest in teaching and/or research rather than exploiting their skills to maximize individual financial return (which is presumably more attractive to conservatives).
Another possibility is that the increasing shift of Republicans to anti-science positions have alienated the scientists and scholars of academia. I think that is a real phenomenon, and it is difficult to wee how it could not be the case given Republican support for degrading education in Biology in favour of teaching religious views as science, and their campaign against, not just action against global warming, but climate scientists. Never-the-less, the survey was taken in 2004, before opposition to action on global warming became truly rabid, and before the success of Tea Party candidates in 2010 made being realistic about climate science the death knell of a Republican politician's career.
Alternatively, left leaning people may be more attracted to the type of reasoning needed for academic success - specifically the ability for nuanced reasoning and suspended belief, a theory for which there is some independent evidence. This should not be mistaken for a theory that conservatives are less intelligent, there being many different ways of being intelligent. However, that is the third possibility; and as conservative academics have certainly been willing to entertain an equivalent hypothesis with regard to race of far less reliable evidence, it is not a theory that should just automatically be dismissed.
(Note, this post is a heavily edited version of a comment from "... And Then There's Physics", although that in turn was the result of my intending to blog on Cardiff and Klein for some time now.)