(C)CES Data

The data is all from the (C)CES. They have been asking, since 2009, how liberal or conservative different candidates are using a 7 point likert scale question. Here I provide some useful (maybe?) details about the data.

Candidates

Year
Governor
Representative
Senator
Incumbent Candidate Incumbent Former Incumbent Candidate
2010 50 841 35 - 63 73
2011 50 - 435 - 100 -
2012 50 842 408 393 78 66
2013 50 - 433 - 100 -
2014 50 836 57 - 73 71
2015 50 - 435 - 100 -
2016 50 845 77 - 78 68
2017 50 - 435 - 100 -
2018 50 845 67 - 68 70
2019 50 - 432 - 100 -
2020 49 857 43 - 69 68
2021 50 - 433 - 100 -
2022 50 856 180 - 73 68
2023 50 - 434 - 100 -
2024 50 - 433 - 100 -

Bridging

A critical need in A-M scaling is for the raters to rate entities that others rate, these are sometimes referred to as “bridges”. Bridges help to pin down the common scale, making ratings comparable. In the C/CES data there are a set of people/entities that everyone rates but these groups change over time. In most cases ths includes the President and the two major political parties.

Year Democratic Party Republican Party Tea Party Movement Supreme Court House of Representatives Senate Mitt Romney Jeb Bush Ted Cruz Rand Paul Hillary Clinton Donald Trump Merrick Garland Joe Biden Kamala Harris Barack Obama
2010 51,096 50,864 46,517 - - - - - - - - - - - - 51392
2011 18,052 17,985 16,708 16,575 - - - - - - - - - - - 18178
2012 67,058 66,292 62,939 62,946 17,203 17,359 64,710 - - - - - - - - 67694
2013 14,329 14,256 13,039 13,296 - - 13,931 - - - - - - - - 14432
2014 56,586 56,446 52,023 53,009 8,621 8,637 - 40,757 34,864 37,064 48,553 - - - - 58029
2015 11,994 11,863 - 10,954 - - - - - - - - - - - 12045
2016 56,293 55,591 - 52,375 - - - - - - 57,072 49,490 5,501 - - 57416
2017 15,698 15,555 - 14,080 - - - - - - - 13,944 - - - -
2018 53,726 53,409 - 50,924 - - - - - - - 49,826 - - - -
2019 15,957 15,755 - 14,543 - - - - - - - 14,311 - - - -
2020 53,017 52,347 - - - - - - - - - 48,380 - 53,037 - -
2021 22,834 22,395 - - - - - - - - - 20,770 - 22,972 - -
2022 53,242 52,333 - 49,397 - - - - - - - 49,301 - 53,259 - -
2023 21,371 21,176 - 20,100 - - - - - - - 20,341 - 21,508 - -
2024 54,403 53,379 - 50,943 - - - - - - - 50,376 - 54,567 54,678 -

Panel Data

There are several years where the CES has included a panel of voters across multiple survey years. In this case I have opted to ignore the panel aspect and treat them as if they are different people. This is likely throwing out some information but it seemed like the best option. There are three ways to approach this:

  1. Assume that voters have not changed at all across the survey panels. For this choice there would be one \(\beta\) and one \(\alpha\) for each voter which would be used repeatedly across panels. This assumes that they have not change their own political views in this time period, having roughly the same understanding of liberal-conservative across the time period. I thought this assumption was overly restrictive.

  2. Use some sort of pooling model where each year’s scores would help inform the other year’s, but not entirely constrain them. This could be implemented with some sort of hierarchical prior on them or even a dynamic prior over time. Although I think this is interesting, the computational complexity concerned me.

  3. Treat them as all independent. This, as I said, ignores potentially useful information but was feasible.

I might return to try 2 at some point, we will see what the future holds.