The Collapse of the Voting Structure — Possible Big Trouble Ahead

Donald Trump will be sworn in as President of the United States next Friday, 20 January 2017. This surprised most analysts including ourselves! We expected to be writing posts during 2017 about splits in the Republican Party due to President Hillary Clinton’s pressing a legislative program that would have been attractive to enough Republicans to have caused serious conflict in the Republican caucuses. Instead we have a President who is both a real estate tycoon and a television entertainer. What his policy views are in many areas are opaque (to say the least).

What we find alarming is the unprecedented collapse of the long-term structure of Congressional Voting during the past 20 years. Contrary to what many scholars say when they cite our book, Congress: A Political-Economic History of Roll Call Voting (1997, New York: Oxford University Press), Poole and Rosenthal DO NOT CLAIM that voting in Congress is largely one-dimensional. Rather Poole and Rosenthal show that a two-dimensional dynamic spatial model is the best fitting model for Congresses 1 – 99.

What has happened in the past 20 years is that the second dimension of Congressional voting has slowly evaporated. As late as the 1990s the second dimension picked up differences within each of the parties over abortion, gun rights, and other social or lifestyle issues. For example, on the vexed issue of abortion each Party had a pro-choice and a pro-life faction. Hence, roll call votes on Abortion often cut through the parties along the second dimension. The same was true for gun control (see the spatial maps in this post). Hare and Poole show the second dimension disappearing in a variety of issue areas in this analysis.

The two figures below show that the extraordinary divisiveness that has marked American Politics since November 2000 has resulted in Congressional voting to collapse into a one dimensional near Parliamentary voting structure; that is, the parties are very unified as shown by Party Unity Scores. The first graph shows the correct classification for each House in 10, 2, and 1 dimensions using Optimal Classification. Note the dramatic convergence of all three measures during the past 20 years. This shows that almost every issue is voted along “liberal-conservative” (it is hard to make sense what this dimension means any more!) lines. Furthermore, no other period in American history shows this pattern.

Click image to enlarge

This graph shows the Aggregate Proportional Reduction in Error (APRE) ([{Minority Vote on a RC – Classification Error}/Minority Vote on a RC], summed over all the scaled roll calls). The APRE controls for the margins of the roll calls. The same pattern of collapse is shown here as well.

Click image to enlarge

Donald Trump will be the third consecutive President who is widely disliked by members of the opposite Party. Indeed, Trump’s personality coupled with the extraordinary party unity within each party will mean that American Politics will enter a phase that has never been seen before. We hope things do not melt down but we would not bet our mortgages on it!

The End of the 114th Congress

The 114th Congress has finally ended. The most polarized Congress since the early 20th Century and one where almost all issues have been drawn into the first dimension [we will address dimensionality in a future post; the current period is unique in American history].

In our last post cited above, we, like most other election watchers, assumed that Secretary Clinton was on a path to a certain victory over Donald Trump. This post would have dealt with what we anticipated to be certain splits within the Republican caucus of the House (we expected the Senate to be 50-50 or 49-51 in favor of the Democrats). All that has changed.

However, it is hard to see how some of President-Elect Trump’s policy positions can be reconciled with the views of the Republican Caucus in either Chamber. For example, in this blog post by Sam Quinones he discusses how the opioid epidemic is correlated with Donald Trump’s Performance in the Election [the analysis in the linked PDF is by Shannon Monnat of Penn State University]. As Quinones notes, these areas that strongly supported Trump will require “massive investment in drug treatment before they can be great again.” We called attention to Quinones’ work in a post early this year and in this op-ed by CDC Chief Thomas Friedman he discusses the seriousness of this epidemic.

Below is a smoothed histogram of the 114th House and Senate using our Constant Space DW-NOMINATE Scores. Note the gap between McConnell (R-KY) and Ryan (R-WI) and the gap between Pelosi (D-CA) and Schumer (D-NY). Add in the filibuster requirement of 60 votes on legislation and it is hard to see how Trump can pass meaningful social policy to help those areas that strongly supported him. Trump may be more successful with taxes and deregulation that would help his supporters but that remains to be seen. The 115th Congress will not be boring.

Click image to enlarge

Alpha-NOMINATE applied to the 114th House (12 September 2016)

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 114th House. There have been 1200 total votes in the House as of the Thanksgiving recess of which 1039 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 0.99957 with a standard deviation of 0.000394 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. The respective Party leaders are near the modes of the two Parties.

Click image to enlarge

The next five plots show the estimated ideal points for the 436 scalable Representatives along with their 95% Credible Intervals. On the left, Representative Grijalva (D-AZ) is located at -2.376. His 95% credible interval runs from -2.577 to -2.182. The five Republicans on the right end are Duncan (R TN-2) at 2.086 (1.84 – 2.30), Huelskamp (R KS-1) at 2.14 (2.37 – 1.93), Massie (R KY-4) at 3.78 (3.94 – 3.61), Amash (R MI-3) at 4.03 (3.89 – 4.17), and Jones (R NC-3) at 8.845 (8.39 – 9.29).

Walter Jones is also the most extreme member of the 113th House. Indeed, many of the more extreme members of the Republican caucus continued into the 114th. With the volatile issue of Planned Parenthood funding holding up the funding for fighting the Zika virus it may prove difficult for Congress to pass a Continuing Resolution that would fund the Government past the upcoming Presidential election.

Click image to enlarge

Click image to enlarge

Click image to enlarge

Click image to enlarge

Click image to enlarge


Law and Order in the 2016 Election

Below we show voters’ feeling thermometer ratings of the police and Black Lives Matter (BLM) using data from the 2015 Cooperative Congressional Election Study. We find considerable polarization between Democratic and Republican respondents’ evaluations of these two groups (the feeling thermometer scores run between 0 [very cold] and 100 [very warm]). Nearly all Republicans are clustered in the upper left: giving police high ratings and BLM low ratings. On the other hand, Democrats are dispersed throughout the space. Democrats have nearly identical mean ratings of the police (56.1) and BLM (55.8).

Click image to enlarge

The 2016 election is reminiscent of the 1968 election in the salience of law and order issues. Using feeling thermometer data from the 1968 American National Election Study, Bayesian multidimensional scaling finds a stark law and order divide running from Humphrey voters on the left to Wallace voters on the right, with Nixon voters in between:

Click image to enlarge

We see the same divide on a question specifically about how best to deal with the problems of rioting and urban unrest. The urban unrest question in the 1968 ANES asked respondents where they and the candidates stood on a seven-point scale ranging from “solve problems of poverty and unemployment” on the left to “use all available force” on the right. Here again, Wallace was able to pick up a substantial amount of voters most in favor of using force to deal with urban unrest.

Click image to enlarge

Ideological Perceptions of Pence and Kaine

Below we update an ongoing project that uses Bayesian Aldrich-McKelvey Scaling to estimate voters’ ideological perceptions of political figures on the 2016 stage. Because the 2014 Cooperative Congressional Election Study asked 50,000+ respondents to place national political figures (like President Obama) and state-level figures like their governor and senators on a liberal-conservative scale, we can use the Bayesian Aldrich-McKelvey scaling procedure to estimate their positions in a common ideological space. We can also estimate credible intervals (the Bayesian analogue of confidence intervals) that reflect uncertainty in the recovered scores.

The survey respondents seemed to do a good job of placing these figures on the liberal-conservative scale. Bernie Sanders is the furthest left, while Ted Cruz is the furthest right. Tim Kaine is the most centrist Democrat, while Chris Christie and Jeb Bush are the most centrist Republicans. Mike Pence has virtually the same score as Marco Rubio: to the right of moderate Republicans, but not as far to the right as Cruz and the Tea Party.

Also noteworthy is that the Supreme Court is nearly dead center in these scores (based on 2014 data). We wait to see if the Court will be placed further leftward by 2016 CCES respondents.

Click image to enlarge

Scaling Voters, Candidates, and Groups with Feeling Thermometers

Below we use Bakker and Poole’s Bayesian multidimensional scaling method to scale voters, candidates, and groups together using respondents’ feeling thermometer ratings. Specifically, the UGA module of the 2015 Cooperative Congressional Election Study asked 1,000 respondents to rate how warmly or coldly they felt about a series of candidates and groups (e.g., investment bankers, Muslims, Evangelical Christians, etc.).

Based on these thermometer ratings, we can jointly scale voters and these candidates/groups such that greater distances correspond to “colder” ratings and smaller distances correspond to “warmer” ratings. The main advantage of the Bakker-Poole method is that it avoids pushing candidates/groups too far to the edges of the space. That is, there are some voters to the left of Sanders and the right of Trump–as a general rule, candidates shouldn’t be the most extreme points in the space. As a technical matter, this is because Bakker-Poole MDS uses the log-normal distribution to model the thermometer ratings, which reflects the intuition that smaller distances should have smaller error variances.

The results are shown below. Because we are using a Bayesian approach, we can easily extract measures of uncertainty for the point locations (Jacoby and Armstrong have also developed a bootstrapping approach to estimate uncertainty intervals for MDS solutions). Accordingly, we also plot the 95% credible intervals for Donald Trump, Marco Rubio, Jeb Bush, Bernie Sanders, Hillary Clinton, and Pope Francis.

Several things stand out. First, there is a clear liberal-conservative division not only among candidates and partisans, but also social groups. There are two distinct clusters of Democratic and Republican candidates, voters, and groups.

That being said, there are also internal divisions within each of the party clusters. Both Clintons and President Obama are highest on the second dimension, while we find Sanders and several liberal groups lower on the second dimension. Socialists are placed the furthest left of all groups. On the Republican side, Trump and Jeb Bush anchor opposite ends of the second dimension. As we would expect, more Republicans have high second dimension scores (closer to Trump) than low second dimension scores (closer to Jeb Bush).

There is more uncertainty associated with the locations of the Republican candidates than with Hillary Clinton or Bernie Sanders. Hillary Clinton is also closer to the interior of the Democratic Party than Trump is to the interior of the Republican Party.

Click image to enlarge

The Politics of Firearms Control

The first serious effort by the federal government to control firearms was The National Firearms Act of 1934 (for a detailed summary, see Anthony Madonna’s summary of the legislative history of the 1934 NFA). The bill was passed on 18 June 1934. The Act taxed the sale of machine guns, sawed off shotguns or rifles, and silencers. The NFA was a response to the notorious gangsters and bank robbers of the era — for example, John Dillinger, Bonnie and Clyde, and Pretty Boy Floyd (all were shot dead by law enforcement officers in 1934). The plague of gangsters who robbed banks and then fled across state lines led to Congress passing The Federal Bank Robbery Act of 1934 on 14 May 1934. This made it a federal crime to rob any national bank or state member bank of the Federal Reserve. This act allowed the FBI aided by local law enforcement officials to pursue bank robbers across state lines. The bill was passed by both Chambers by voice vote.

The debate over the NFA prefigures some of the debate taking place today. The decision to tax and register machine guns seemed to implicitly assume that individuals had the right to own firearms under the second amendment. Certainly firearm ownership was relatively widespread in the 19th Century in rural areas of the United States and Congress made no effort to limit their ownership. The wording of the second amendment is hotly contested in the 5-4 District of Columbia et al. v Heller (2008) decision. The five justices in the majority held that individuals had the right to own a personal firearm whereas the four justices in the minority held that this right was linked to membership in a (State) militia. Needless to say, each side constructed their arguments around different interpretations of the Founders’ intent and early history of the Republic.

The next significant Act that regulated firearms was the Gun Control Act of 1968. It was a response to the murders of President John F. Kennedy (22 November 1963), Dr. Martin Luther King (4 April 1968), and Robert F. Kennedy (6 June 1968 — he was shot by Sirhan Sirhan on 5 June 1968). President Kennedy was murdered by Lee Harvey Oswald with a bolt action rifle. Dr. King was murdered by James Earl Ray with a bolt action rifle. And Robert Kennedy was murdered by Sirhan Sirhan with a 22 caliber pistol.

The 1968 Act finally banned mail order sales of rifles and shotguns and prohibiting most felons, drug users and people found mentally incompetent from buying guns. The key votes shown below were lopsided with majorities of both political parties supporting the law.

Click image to enlarge

Click image to enlarge

President Johnson, echoing some of the debate of the 1934 NFA, when he signed the 1968 Act stated: I asked for the national registration of all guns and the licensing of those who carry those guns. Clearly he felt that national registration and licensing was Constitutional and did not violate the second Amendment.

The next significant piece of legislation on firearms control was The Brady Handgun Violence Prevention Act passed by Congress in 1993 and backed by former President Ronald Reagan. The Brady Bill required that background checks be conducted on individuals before a firearm may be purchased from a federally licensed dealer, manufacturer or importer with a five day waiting period before the purchase could be completed. The National Rifle Association opposed the law and managed to get a provision inserted that mandated the National Instant Criminal Background Check System be put into effect by 1998. This system is still in use.

The Brady Bill split both political parties internally. In the House the old (and dying) Conservative Coalition was against the Act but it still passed by a comfortable margin.

Click image to enlarge

Click image to enlarge

Almost at the same time as the Brady Act was being considered a ban on Assault Weapons was also being debated. The attention to Assault Weapons (basically, military style semi-automatic rifles and pistols) began with a 1989 school yard shooting in Stockton, California and later a 1993 shooting in San Francisco, California. Both shooters appeared to have been insane and committed suicide rather than be captured. The Stockton shooting led President George H.W. Bush to ban all imports of Assault Weapons into the United States in 1989.

The votes on the Assault Weapons ban were closer than those for the Brady Bill. Once again, the votes split both political parties internally with the vote in the House being very close. In a preview of what was to come the Senate Democrats were more united in favor of Gun Control than House Democrats. The political parties were beginning to move to a polarized position on the issue of Gun Control.

Click image to enlarge

Click image to enlarge

Next up was a push by firearms supporters for the protection of gun makers from civil lawsuits over the use of their guns in crimes. This was the Protection of Lawful Commerce in Arms Act. The basic idea was to protect manufacturers from being sued because someone criminally used a firearm. As Senator Sanders (I-VT) put it during the 2016 Democratic Presidential Primary Campaign: “If somebody has a gun and it falls into the hands of a murderer and the murderer kills somebody with a gun, do you hold the gun manufacturer responsible? Not any more than you would hold a hammer company responsible if somebody beats somebody over the head with a hammer.”

The votes are similar to the Assault Weapons Ban only now the Republican Party is now almost completely pro-firearm while the Democratic Party is split. The high PRE values indicate that the issue is become more divisive and set the stage for the battles during the Obama Administration.

Click image to enlarge

Click image to enlarge

During the Administration of President Obama a number of high profile shootings has raised the salience of gun control to a very high level. These shootings include the mass casualty shooting in a movie theater in Aurora, Colorado, on 20 July 2012 (12 dead, the shooter, James Holmes, was later ruled to be insane), the Sandy Hook Elementary School shooting in Newtown, Connecticut on 14 December 2012 (27 dead and the shooter, Adam Lanza, committed suicide [he had murdered his mother before he went to the school]), and most recently, the Pulse Nightclub shooting in Orlando, Florida on 12 June 2016 (49 dead and the shooter, Omar Mateen, was shot dead by police).

In all three shootings semi-automatic weapons were legally obtained (in the Sandy Hook case, obtained by his Mother) and used by the shooters. In the cases of James Holmes and Omar Mateen a better background check system should have prevented them from buying the weapons they used in the shootings. Holmes had been seen by a Psychiatrist at the University of Coloradoand evidently “fantasized about killing a lot of people.” Mateen was known to the FBI but this evidently was not enough to flag him during a firearms background check. Consequently, recent legislative action in Congress has focussed on beefing up the background check system.

So far action has only occurred in the Senate. On Monday, 20 June the Senate voted on the Republican version of a beefed up background check system that would encourage states to share mental health records as part of the system. The Senate also voted on the Democrat version that would require a background check for most sales or transfers of guns.

Unfortunately, from 1934 when firearms control was unanimously supported by both parties, the issue has now fallen into the vortex of partisan polarization. Almost all sensible people agree that, at the least, the background check system should be considerably expanded to include persons on the “no fly list” and include records on individuals who are clearly insane. The devil is in the details. In the past when there were a substantial number of moderates in both parties this would be an easy question to resolve. Some sort of semi-judicial system should be set up to decide if someone was unfairly denied the right to purchase a firearm. But 2016 is an election year and gun control has become a deeply partisan issue. It is possible that there may be a compromise but that seems to be a long shot as things stand at the moment.

Click image to enlarge

Click image to enlarge

Ideological Positions of Clinton and Sanders Voters

In this post, we look at the ideological positions of Clinton and Sanders supporters among non-Republican respondents to the pilot study of the 2016 American National Election Study. We look at both the raw and weighted responses (in which survey weights are applied to approximate the demographic makeup of the population).

Even with weighting applied, non-Republican respondents who support Sanders in the Democratic primary are to the left of Clinton on most issues, especially the issue priority questions (that is, Sanders voters are more likely to say that income inequality and the environment are among their top issues). These issue salience measures have been shown to be influential in voters’ evaluations of candidates.

In the last two graphs, we also use the Aldrich-McKelvey scaling procedure (detailed in an recent post) to get a handle on the ideological positions of non-Republican Clinton and Sanders supporters. This method uses respondents’ placements of themselves and parties/candidates on the liberal-conservative scale to produce ideal point estimates that adjust for biases (or differential item-functioning) in the way respondents use these scales. Again, both with and without weighting, non-Republican Sanders supporters seem to be somewhat more liberal than Clinton supporters. These differences between the two camps aren’t massive, but may nonetheless be signaling something of an ideological split in the Democratic party.

Click images to enlarge

Issue Priorities, Partisanship, and Crossover Voting in 2016

In this post, we look at the relationship between issue salience, candidate support, and partisanship vary by candidate support using data from the pilot study of the 2016 American National Election Study. This is part of a larger project that examines how different types of Democrats and Republicans vary in their issue priorities. In addition, issue salience may help to get a better handle on crossover voting (Democrats voting for Trump/Republicans voting for Clinton) in the 2016 election.

Below, we plot the percentages of each candidate’s supporters (whether they answered Clinton, Trump, or other/not vote in the general election matchup question) naming the given issue as one of their top four most personally important issues.

We show just Democrats and independents in the first plot, and isolate Republicans and independents in the second plot. We include independents in both groups because the sample sizes for Trump Democrats and (especially) Clinton Republicans is negligible (there are 24 Democrats who support Trump and 4 Republicans who support Clinton). Combining independents with both groups seems to reinforce the patterns when isolating respondents by partisanship (we provide these plots at the bottom of the post), but provides more leverage for the analysis.

The results indicate that crossover voters are indeed more likely to name issues “owned” by the opposite party as important. Issues like immigration, the national debt, and terrorism are more salient to respondents (including Democrats/independents) who support Trump. Conversely, the issues of environmental protection, health care, and income inequality are more salient to Clinton supporters (including Republicans/independents). It doesn’t appear that issue salience is of much help in explaining which respondents say they would vote for another candidate or not vote in a Clinton/Trump matchup—these respondents simply aren’t distinctive in their issue priorities. We thought that the other/not vote Republicans would be more likely to prioritize social issues (abortion and morality), but this doesn’t appear to be the case.

Certainly, some of this is a projection effect, but, especially in a year in which ideology plays such a murky role, issue salience can pick up the slack in modeling vote choice and crossover voting.

Click images to enlarge

Below are show separate plots for Democratic, independent, and Republican respondents. The general findings are consistent with the above graphs, but (especially when it comes to Republicans who support Clinton), the sample sizes are quite small.

Click images to enlarge

Ideological Perceptions of the 2016 Presidential Candidates

Below, we use Aldrich-McKelvey scaling (see here and here for background) to analyze voters’ ideological perceptions of the 2016 presidential candidates and other political figures. The 2016 pilot study of the American National Election Study asked respondents to place themselves and each of these figures on a seven-point ideological scale ranging from “extremely liberal” to “extremely conservative.” The Aldrich-McKelvey scaling procedure allows us to recover bias-corrected estimates of the respondents and candidates on the underlying ideological dimension.

The estimated scores (plotted below) show that three Democratic figures (the Democratic Party, President Obama, and Secretary Clinton) are ideologically clustered together. Clinton is a bit closer to the center than the other two, but not by much. On the other hand, there is considerable heterogeneity in ideological perceptions of the four Republican stimuli: Senator Rubio, Donald Trump, the Republican Party, Senator Cruz. It’s probably not surprising that Rubio is perceived to be the most moderate and Cruz is perceived to be the most conservative of the four.

What is noteworthy is that Trump is placed at nearly the same spot as Rubio. Trump, however, has the the greatest amount of uncertainty associated with his ideological score (as a technical note, we estimate 95% confidence intervals for the Aldrich-McKelvey scores using 1,000 bootstrap replications, as described in Chapter 3 of our book on estimating spatial models).

This is equivalent to saying that respondents differ the most in where they place Trump on the ideological scale. The width of Trump’s confidence interval is about twice that of Clinton’s, for instance. This uncertainty could be a factor in the 2016 race, as some political science research suggests that voters reward candidate ambiguity (see also here).

Click image to enlarge

We wondered if voter uncertainty about Trump’s ideological position was being driven by a divide among self-identified conservative respondents; that is, between those who embrace Trump and believe he’s a conservative, and those (e.g., #NeverTrumpers) who doubt Trump’s conservative credentials. To look into this possibility, we plotted the mean placements of Trump and Clinton by respondents’ ideological self-identifications. “1” indicates extremely liberal, “2” indicates liberal, and so on until “7” for extremely conservative. Respondents are sorted in this way on on the vertical axis of the graph below.

For each ideological category of respondents, mean placements of Clinton and Trump are shown along the same seven-point liberal-conservative scale on the horizontal axis. For example, the most conservative respondents (self-identified “7”‘s or extremely conservative) place Clinton furthest to the left. The gray bars represent variation in the corresponding group’s ideological placements of Clinton and Trump.

Interesting, across the ideological categories, respondents are pretty evenly uncertain about Trump’s position on the liberal-conservative scale. On the whole, self-identified conservatives do view Trump as somewhat more moderate than do self-identified liberals (evidence of what is known as interpersonal incomparability or differential item-functioning). But, it is self-identified moderates who place Trump closest to the center (and have the least variation in their Trump placements).

Click image to enlarge