Multidimensional Scaling of Feeling Thermometers from the 2015 CCES and 2016 ANES Pilot Study

Below, we use a multidimensional scaling method (metric unfolding) to analyze feeling thermometers data from two recent national public opinion surveys: the 2015 Cooperative Congressional Election Study (CCES) and the 2016 American National Election Study (ANES) Pilot Study. Both surveys asked respondents to rate their feelings towards candidates and groups (for example, Donald Trump, Hillary Clinton, and Muslims) on a 100 point scale. We transform the thermometer ratings into distances and produce a spatial map of the results, much as we would to produce a map of cities from on a spreadsheet of driving distances between the cities.

The results are shown below, with the survey respondents marked as “D”, “R”, and “I” based on their party identification. In both plots, the first dimension (the horizontal axis) represents the familiar partisan-ideological divide, separating liberal/Democratic groups and candidates from conservative/Republican groups and candidates. Groups like scientists and college professors are placed in the center-left, while groups like the police are placed center-right.

We suspect that the second dimension is tapping into establishment vs. outsider divisions in both parties: between Hillary Clinton and Bernie Sanders among Democrats, and between Donald Trump and the other candidates (particularly Jeb Bush) among Republicans. Trump is preferred by the larger cluster of Republican voters in the top-right quadrant of both plots, while Bush is preferred by the smaller clusters of Republicans in the bottom-right quadrants. Cruz and Rubio are both somewhere in between.

Within the parties, respondents’ relative preferences between the candidates do not appear to follow a traditional liberal-conservative divide, but are structured along a separate dimension—perhaps involving establishment vs. outsider attitudes and preferences on cross-cutting issues like free trade.

Click images to enlarge

The Heroin Epidemic

This past Wednesday, 2 March 2016, there was a procedural vote in the Senate to waive budget caps in order to appropriate more funds to fight the epidemic of Heroin and other Opioids (just search Google News for countless recent articles across the country on this). Sam Quinones in his book Dreamland details the origins of the recent epidemic in the widespread use of Oxycodone (OxyContin is the time released form of Oxycodone) and Black Tar Heroin from Mexico. The Opioid epidemic became an issue in the recent New Hampshire Primary but it is a nationwide problem. Atlanta’s Channel 11 television news just did a four part series on how widespread Heroin is in Atlanta’s affluent northern suburbs where packages of heroin are delivered to the front doors of users (the user leaves money under a doormat and the deliveryman takes the money and leaves the heroin).

Below we show the procedural vote on more funds to fight the Opioid epidemic using our Constant-Space DW-NOMINATE Scores. Five Republicans voted “Yea” — Ayotte (R-NH), Collins (R-ME), Graham (R-SC), Kirk (R-IL), and Portman (R-OH). The cutting line goes through the lower left portion of the Republican Party and there are only four “errors” on the roll call with a PRE of 0.91.

Click image to enlarge

Portman’s vote for addition funding is significant. Quinones starts his investigation of the twin epidemics in Portsmouth, Ohio, on the Ohio river. Southern Ohio has been hit hard by the Opioid epidemic and Governor Kasich expanded Medicaid under the ACA in part because of the impact of Oxycodone and Heroin in Ohio. His argument is that treatment and helping the addicted to regain their independence and hold jobs is better than incarceration. Hopefully his point of view will prevail.

The Francis Vote?

Even before this week, Republican voters were split over Pope Francis and Donald Trump. In the 2015 Cooperative Congressional Election Study, we asked respondents to rate a variety of political groups and figures (including Pope Francis and Donald Trump) on a 100-point “feeling thermometer,” with higher values corresponding to more favorable evaluations.

Below, we divide the roughly 250 Republican and Republican-leaning respondents in the survey into two groups: those who rate Trump higher than Francis, and those who rate Francis higher than Trump. We then split respondents by whether they rate Trump, Cruz, or Rubio most highly on the feeling thermometers.

Two things are clearly seen in the plot: most Republicans (about 2/3) rated Trump more highly than Francis (this survey was fielded in late 2015); and Trump (naturally) wins the Trump greater than Francis group, while Rubio is the big winner in the Francis greater than Trump group.

Click image to enlarge

Next, we look at which types of voters comprise the two groups in the GOP. We specify a binary probit regression model in which the response variable is coded 1 if the respondent prefers Francis to Trump, and 0 if the respondent prefers Trump to Francis. Below we present the effects of a series of variables (all coded to range between 0 and 1) in terms of the change in predicted probability of a Republican respondent being in the pro-Francis group.

Not surprisingly, conservative immigration attitudes most strongly predict pro-Trump dispositions, while education increases the probability of preferring Pope Francis to Trump.

Catholic identity has perhaps the most intriguing effect: whether or not Republican Catholics are more likely to side with Pope Francis over Donald Trump depends on level of religious activity and involvement. Those who are most active (in the top half of a religiosity scale comprised of level of prayer, church attendance, and personal religious salience) are more likely (about 35% more likely) than other Republicans (including less religiously active Catholic Republicans) to rate Pope Francis more highly than Donald Trump.

This might seem a fairly obvious finding (more active Catholics are more strongly committed to the Pope), but Pope Francis has also stood out in his appeal to non- and lapsed Catholics (the so-called Francis Effect). However, there is no apparent Francis effect (at least as it involves feelings vis-a-vis Trump) among Republican Catholics.

Click image to enlarge

More on Polarization Through the 114th

A number of people have requested that we provide “regular” DW-NOMINATE scores for the House and Senate using the dynamic two-dimensional model through calendar 2015. We have computed these and will post them on our data for friends page this weekend.

Below we show the polarization series for the House and Senate separate scalings. This graph is almost identical to the one in our post a few weeks ago using the Weekly CS DW-NOMINATE scores. As we noted in that post, the impact of the last three elections — 2010, 2012, and 2014 — on the Republican Party seems to have been the large influx of “Ted Cruz” Republicans. Polarization has jumped sharply from the first two years of President Obama’s first term. Polarization in the House may have leveled off (see below) but the Senate increase is much larger than the House.

Click image to enlarge

The figures below show the means of the two parties in each Chamber on both dimensions over time. The House Republicans actually moderated very slightly in the first session of the 114th while the House Democrats show absolutely no difference between North and South and the two have converged. The slight uptick in polarization in the House is due to the leftward movement of the Democrats being slightly greater than the leftward movement of the Republicans.

The Senate for the most part tracks the House on the first dimension. The main difference is that there are still three Senators from the South (the eleven states of the Confederacy plus Kentucky and Oklahoma [CQ’s definition]). Those three [Nelson (D-FL), Kaine (D-VA), and Warner (D-VA)] are more moderate than their Northern counterparts. Nonetheless, the Senate Democrats are moving to the left and the Senate Republicans are moving to the right thereby increasing polarization in the Senate at a faster rate than in the House.

Click image to enlarge

The Second Dimension at one time picked up conflict over Civil Rights for African Americans but, beginning in the early 1970s until the early 2000s “Social Issues” such as abortion and gun control (see Poole and Rosenthal, 1997; McCarty, Poole, and Rosenthal, February 2016). The Second Dimension has faded to insignificance but could make a come back with a possible split in the Republican Party during the 2016 elections.

Click image to enlarge

Click image to enlarge

Click image to enlarge

The Divisions in the House Republican Party

(Clarifications made 0020UCT 9 January 2016)

To recap: As we discussed in earlier posts, alpha-NOMINATE is a new form of NOMINATE that is fully Bayesian and is meant to replace W-NOMINATE which is now about 33 years old (the multidimensional version, written by Nolan McCarty and Keith Poole, is almost 25 years old). NOMINATE was designed by Keith Poole and Howard Rosenthal during 1982-1983. It used a random utility model with a Gaussian deterministic utility function (see pages 14 – 15 of the linked 1983 paper) and logistic error (random draws from the log of the inverse exponential). The Gaussian deterministic utility function is able to capture non-voting due to indifference and alienation.

Alpha-NOMINATE is a mixture model in which legislators’ utility functions are allowed to be a mixture of the two most commonly assumed utility functions: the quadratic function and the Gaussian function assumed by NOMINATE. The “Alpha” is a parameter estimated by Alpha-NOMINATE that varies from 0 (Quadratic Utility) to 1 (Gaussian Utility). Hence, in one dimension with Alpha = 0, Alpha-NOMINATE is identical to the popular IRT model. Thus Alpha-NOMINATE can actually test whether or not legislators’ utility functions are Quadratic or Gaussian.

Below we apply Alpha-NOMINATE to the first Session of the 114th House. There were 705 total votes in the during the first Session of which 614 are scalable (at least 2.5% in the minority; that is, votes that are 97-3 to 50-50). We used the R version of Alpha-NOMINATE to perform the analysis. We used 4000 samples from a slice sampler in one dimension with a burn-in of 1000. The first graph shows the Trace and Density plots for alpha.

Click image to enlarge

The mean of alpha is 0.99986 with a standard deviation of 0.000014 strongly indicating that the Representatives’ utility functions were Gaussian.

Below is a smoothed histogram of the 3000 configurations after burn-in. The divide between Democrats and Republicans is a very deep one.

Click image to enlarge

Of more interest, however, are the clear divisions in the Republican Party shown in the smoothed histogram. Ryan Lizza in a recent article in the New Yorker, 14 December 2015, discusses these divisions in a particularly lucid fashion. He quotes Charlie Dent (R-PA, CS DWNOM Score of 0.243 on the first dimension) that there are 70-100 Republicans like himself that will vote with the Democrats to pass the Omnibus bills to keep the government running, 70-80 “hope yes, vote no’ Republicans, who voted against those bills but secretly hoped they would pass; and the the 40-60 members of the rejectionist wing, dominated by the Freedom Caucus, who voted against everything and considered government shutdowns a routine part of negotiating with Obama” (p. 37). The Rejectionist wing’s de facto leader is Ted Cruz (R-TX, CS DWNOM Score of 0.975).

Fueled by talk radio the Rejectionists would shut the government down until their demands are met no matter what the cost. Essentially, the House Republicans are being held hostage by these extreme True Believers. Cruz is running a sophisticated campaign for the Republican nomination and it is conceivable that he could win. Cruz suffers from the same delusion as Barry Goldwater in 1964. Out “there” are millions of dissatisfied voters who suddenly will flock to the polls when a “true” conservative is nominated (see, Converse, Clausen and Miller, 1965. “Electoral Myth and Reality: The 1964 Election.” American Political Science Review, Vol. 59, No. 2 (Jun., 1965), pp. 321-336). The first page of that 1965 article is eerie because it could easily be read as a description of the current debate within the Republican Party.

The reality is that there are not millions of Republicans who will flock to the polls to vote for either Ted Cruz (the most despised Man in the Senate) or Donald Trump. As Bret Stephens advises Wall Street Journal readers in his Global View column on 22 December 2015 “Let’s Elect Hillary Now,” he sees the Republican Party losing on the scale of 1972. If so, there will be a decisive turn towards European style Social Democracy (i.e., a vast increase in social programs and greatly increased taxation) under Hillary Clinton (CS DWNOM Score of -0.373 as Senator from New York). However, to make things even more complicated, Bernie Sanders (I-VT, CS DWNOM Score of -0.513) may very well upset Clinton in Iowa and win New Hampshire. Clinton is the overwhelming favorite but is widely distrusted by the public (for example, Bill Clinton’s escapades are beginning to resurface — a particularly toxic one is the Juanita Broaddrick rape allegation). Nevertheless, the smart money is on Clinton and a crackup of the Republican Party.

Polarization Continues Through 2015

The first session of the 114th Congress was slightly more polarized than the 113th Congress. However, what is important about the polarization graph below is the impact of the last three elections 2010, 2012, and 2014, on the Republican Party. The large influx of “Ted Cruz” Republicans has caused polarization to jump sharply from the first two years of President Obama’s first term. Polarization in the House may have leveled off (see below) but the Senate increase is much larger than the House. These results are from our final set of Weekly CS DW-NOMINATE Scores where the House and Senate are scaled together and each unique member of Congress receives a single score on the first and second dimensions.

Click image to enlarge

The figure below shows the means of the two parties in each Chamber over time. For the most part the Senate means are more to the interior than the House means (the bright blue and bright red lines) but not always. In the last two elections, the Republican caucus in the Senate has become as conservative as the Republican caucus in the House. In contrast, Democratic Senators are clearly more moderate than their House counterparts.

Click image to enlarge

Given how extreme Republican primary voters have evidently become (the base of the Party), it should be no surprise that a steady movement to the Right by Republicans coupled with the emergence of talk radio and now various social media platforms, that the Presidential nominating process has been “hijacked” by the extremes. In short, we may finally becoming to the end of the polarization that began in the 1970s simply because the Republican Party may fracture in 2016.

The Transportation Bill Finally Passes

Congress finally approved a Transportation Authorization Bill on Thursday, 3 December 2015. The Bill authorizes $305 Billion dollars over a five year period. The $305 Billion would come mostly from the gasoline tax (which has not been raised since 1993) and partly from other “gimmicks” such as raiding the Federal Reserve’s reserve fund. The Bill also authorizes $10 Billion for Amtrak, and it allows Amtrak to pour the profits from the popular Acela Northeast Corridor route back into infrastructure improvements in the corridor (which are very badly needed). The bill also revives the Export-Import Bank and includes a host of other provisions for urban rail transportation and railroad oil tank car safety standards supported by the American Association of Railroads.

Below is the final passage vote in the House. The vote was 350-65 with the 65 “Nays” coming primarily from Conservative Republicans. Nonetheless, the vote has a respectable PRE of 0.32.

Click image to enlarge

In the Senate the bill passed 83-16 (Sanders (I-DE)) did not vote. Here two Democrats, Warren (D-MA) and Carper (D-DE) voted “Nay” along with 14 mostly Conservative Republicans. Note that the cutting line could be moved slightly to the left and reduce the number of errors from 10 to 7. However, CS DW-NOMINATE is a probabilistic model so that moving the cutting line slightly to the left would result in all the probabilities of the “Yea” voters to become slightly smaller. Optimal Classification would position the cutting line to maximize the correct classification because it is not a probabilistic procedure.

Click image to enlarge