Legislator coordinates are probably a function of multiple factors, including ideology, pressure from party organizations, constituent demands, and others. McCarty, Poole, and Rosenthal (2001) demonstrate that parties influence legislator coordinates.
DW-NOMINATE results are sometimes informally described as ideology scores, disregarding the nuances. An older and more neutral term, “ideal point,” is often used to describe the coordinates in more formal research.
The first dimension is the most important dimension in terms of explaining votes, the second dimension is the second most important, and so on.
When applied to US legislatures, the first dimension is usually described as an economic or liberal-conservative dimension, while the second dimension has different meanings at different times.
Aldrich, Montgomery, and Sparks (2014) show that, in the presence of party polarization along multiple dimensions, the main dimension will be a composite of dimensions that split the parties; it’s not necessarily a single political dimension.
Measuring party loyalty in a spatial model requires both a legislator’s coordinates and the coordinates of the party, so that a relative distance can be calculated. But parties aren’t assigned scores by DW-NOMINATE (in fact, parties are not included in the model at all), so DW-NOMINATE scores by themselves are agnostic with regard to party loyalty.
Ted Cruz is a good counterexample to claims that scores reflect party loyalty. Cruz is reported to be widely disliked by his fellow Republican legislators, but, because he disagrees from a more extreme position, DW-NOMINATE places him even farther right than most other Republicans. In this case a more extreme score indicates less party loyalty.
Poole and Rosenthal sometimes speak loosely of a “party loyalty dimension” (2011, 54–55, 1991, 235). A more accurate term would be “party disagreement dimension.”
Sometimes legislators deviate from the spatial model of voting assumed by DW-NOMINATE. In this case their scores will have relatively large standard errors and may not be accurate. This is a limitaton of DW-NOMINATE. In particular, DW-NOMINATE doesn’t include the concept of a protest vote.
Jeff Lewis at Voteview examines this issue in more detail in his articles here and here.
DW-NOMINATE constrains the results so that the mean of each legislator’s scores lies within the unit circle. An extreme legislator with a score that changes over time can move outside of the unit circle for part of his or her career.
Aldrich, John H., Jacob M. Montgomery, and David B. Sparks. 2014. “Polarization and Ideology: Partisan Sources of Low Dimensionality in Scaled Roll Call Analyses.” Political Analysis 22 (4): 435–56. https://doi.org/10.1093/pan/mpt048.
McCarty, Nolan, Keith T. Poole, and Howard Rosenthal. 2001. “The Hunt for Party Discipline in Congress.” American Political Science Review 95 (3): 673–87. https://doi.org/10.1017/S0003055401003069.
Poole, Keith T., and Howard Rosenthal. 1991. “Patterns of Congressional Voting.” American Journal of Political Science 35 (1): 228–78. https://doi.org/10.2307/2111445.
Poole, Keith T., and Howard L. Rosenthal. 2011. Ideology and Congress. Transaction Publishers.