DW-NOMINATE didn't create the buckets "far-left", "left", etc. DW-NOMINATE is just a method of marking plots on a graph [1]. The categories of far-left, left, etc. come from drawing boundaries on that graph and giving everyone inside that region a label. Where to draw these boundaries hugely influences the results. Give someone else task of defining the boundaries and you can easily achieve the completely opposite result: like gerrymandering, you can define boundaries on the DW-NOMINATE chart to fit your desired outcome.
Furthermore, there's only 535 members of Congress, out of 300+ million Americans. DW-NOMINATE only examines a very very narrow segment of the population.
By comparison, the Pew survey was based on directly asking people their positions on various issues and measuring political uniformity. This didn't involve the highly subjective task of drawing regions on a chart and deciding these people are center left, these people are far left, etc. Instead it examined how likely people were to have uniform versus heterodox views. It also didn't exclusively study on politicians, it used representative sampling of the general population.
I think it's pretty funny you're acting like looking at the unrandom sample of elected officials is worse than grabbing assorted randos when trying to study politics. It's even funnier that you're taking people's off-the-cuff opinions given to a stranger stranger more seriously than you're taking the actual political behavior of passing laws. Believe what you like, I suppose, but I have a hard time taking you seriously here.
Pew is one of, if not the most respected polling organization in the US. By comparison, that XKCD chart is all just someone drawing boundaries on a chart. Just make the far left zone larger and shrink the far-right zone and voila you have the opposite result.
Furthermore, there's only 535 members of Congress, out of 300+ million Americans. DW-NOMINATE only examines a very very narrow segment of the population.
By comparison, the Pew survey was based on directly asking people their positions on various issues and measuring political uniformity. This didn't involve the highly subjective task of drawing regions on a chart and deciding these people are center left, these people are far left, etc. Instead it examined how likely people were to have uniform versus heterodox views. It also didn't exclusively study on politicians, it used representative sampling of the general population.
1. https://en.wikipedia.org/wiki/NOMINATE_(scaling_method)#/med...