Following Nate Silver’s analysis of how the issue of domestic surveillance divides the parties internally, below we use Optimal Classification (OC) in R to analyze the 112th House and Senate’s votes to renew the Patriot Act and the FISA (Foreign Intelligence Surveillance Act) Amendments in 2011 and 2012, respectively.
As Mr. Silver rightly notes, voting on these two pieces legislation in both chambers does not conform neatly to a liberal-conservative dimension as do a slew of other issues: tax rates, health care, gay marriage, etc. The Democratic and Republican members of Congress who opposed the extension of these programs do tend to be among the more liberal and conservative (measured by the first dimension) members of their respective caucuses, but members’ second dimension scores are a better predictor of their roll call votes on renewing the Patriot Act and FISA. This is evidenced by the angle of the cutting lines (which divide predicted Yeas from predicted Nays) in the plots below, all of which are more horizontal (separating MCs along the second dimension) than vertical (separating MCs along the first dimension). The PRE (proportional reduction in error) fit statistics are high for all votes, indicating that the cutting lines are effective in modeling voting patterns with the use of both dimensions.
Previous conflicts like civil rights in the mid-twentieth century and bimetallism in the late-nineteenth century emerged as “second dimension” conflicts, meaning that the underlying (first) liberal-conservative dimension is insufficient to explain cleavages over these issues. It is too early too tell whether the second dimension is truly capturing an “establishment vs. outsider” divide over issues like domestic surveillance and the debt ceiling in contemporary congressional voting or merely fitting noise on a special subset of roll call votes, but the evidence so far is suggestive that the second dimension may be real and important.