More fun with special elections

Guest author: Danny Metcalf

 

I promised a few weeks ago to update my earlier post about special elections after results from Montana’s at-large race came in. Obviously, I completely missed that window, but now we have a couple more special elections to talk about.

 

Before I actually get into this project, let’s talk about the elephant in the room, and by “the room,” I mean the congressional office for the representative from Georgia’s 6th district. In a moment, I’m going to talk about how this sequence of special election results is great news for Democrats. That’s going to feel pretty hollow now that Karren Handel (R) won the highest-profile special election of the year (ever?). It seems clear that the way this election result is going to be interpreted in the public consciousness and in Congress, is as a repudiation of claims of Democratic momentum and a sign that Republicans can continue with their policy agenda. There might be real tangible consequences to healthcare, tax reform, and other legislation from an emboldened GOP caucus. There might be strategic consequences by which leaders in both parties rethink where and how campaign dollars are spent. When I write, in a few paragraphs, that Democrats are in a very strong political position, you should understand that I don’t mean anyone should be happy about a loss. On the other hand, just because we all hyped up this race as a binary determinant of the current political climate doesn’t mean that is actually the best way to look at it. The interpretation of the Georgia’s 6th district election might be that Dems have been rebuked, but honestly, that interpretation is dumb.

 

For discussion of what I’m trying to do with this project, see my earlier post, but I’ll give summarize the idea here and then dive into results.

 

General congressional elections in the US consist of 435 distinct races happening in different districts at the same time. Since they are occurring at the same time, it is natural to expect some relationship in how two different districts vote. That is, a good Republican year in Missouri will probably be a good Republican year in Nevada, to the extent that national politics relate to district-level results. On the other hand, two districts have totally different populations. A deep red district in rural Alabama will vote more Republican than a deep blue district in Los Angeles, regardless of the national political environment. Finally there is some variation between seats not captured by partisanship or election year. Two equally deep red districts might have meaningfully different results because of candidate level characteristics, campaign spending, other unobserved factors, or sheer dumb luck*. I therefore model election results happening at the same time as the national partisan environment plus district partisanship plus a random error term.

 

For this project, I am treating 2017 special elections as if they are being produced by a general election happening at the same time and predicting what the rest of the country would look like in that hypothetical election**. The previous post discussed sampling results using election returns from the GA-06 first round and the KS-04 election. Results here will include KS-04, the GA-06 second round, the Montana-at large race, and the SC-05 race. Note that Republican candidates won the KS-04 and MT-AL races fairly comfortably and the GA-06 and SC-05 races more narrowly. It is completely fair to say that those represent good results for Republicans because the actual, real-world consequences are that more Republicans are in Congress. That said, all four of these races are in districts that Republicans should expect to win in a neutral year. The Kansas, Montana, and South Carolina districts only become competitive in national politics in Democrats’ wildest pipe dreams and Republicans’ darkest nightmares. The Georgia seat is a little different. PVI, the method I use to measure partisanship, rates it as R+8, a strongly republican seat, but slightly more moderate than the other three. Other methods that weight the 2016 election more heavily than 2012 will get a distinctly moderate district. At R+8, the actual result of Handel winning by 4 looks very good for Dems. At a more moderate R+2, it’s a bit of a disappointing result. While how you measure partisanship matters for GA-06, the crux of the point is that Democrats are achieving narrow defeats in areas where they should expect to be crushed. That is indicative of a very strong national environment for Dems.

 

OK, onto the simulation results. According to my model, in a general election that produces these four district results, Democrats should expect to take the House 87.4% of the time with a median of 242 seats. There is still quite a bit of uncertainty around these results, which should be unsurprising given that I’m trying to generalize from only four districts. In 1.1% of simulations, Republicans actually gained seats and the 95% credible interval ranges from 200 to 283 Democrats.

 

For those of you who read the earlier post, you’ll notice that all of those numbers are substantially better for Democrats than before the MT-AL race. The reason why is not that MT-AL, SC-05, or the GA-06 runoff were disproportionately good results for Democrats. In fact, they was just about in line with what happened in the earlier two races. GA-06 was a bit worse for Dems and SC-05 was a bit better. Instead, the change basically represents greater confidence in the predictions of the model. These are the sorts of outcomes that we would expect to see in a wave election, where Dems end up with a decent Congressional majority. The model is just a little cautious about predicting a wave from such sparse data. These are also the sorts of outcomes we would expect from the tails of the distribution in a slightly Republican year or a tsunami (what I am arbitrarily choosing to call a ridiculously strong wave election) Democratic year. As more elections come in, it becomes less likely that we’re drawing from the tail of the distribution and more likely that these results are representative.

 

One more comment about GA-06 and SC-05 just to hammer the point home. Democrats convinced themselves they could win in GA-06. They spent heavily there and focused national attention on one special election. Of the four races, so far, it was probably the most winnable for the Democrat, so in some sense the attention in reasonable. In another sense, it was always a long-shot. If you’re a college football fan and your team loses a week-one opener against Alabama in overtime, you’re totally justified feeling upset. You are not justified in giving up on the season. All this race tells us is that Democrats are not riding the sort of historic wave that would completely sweep out every Republican outside of the deep South. More to the point, a 4 point win by the Republican in GA-06 represents a 3.9 point shift towards Democrats from PVI. There are 227 House seats that are R+3.9 or less, meaning even if the Georgia result is more accurate than all the other special elections so far, Dems still have a good chance of taking the House. Also remember that, although it received a fraction of the coverage, there was also a race in South Carolina’s 5th yesterday. In a deeper red district, the Rep won by a smaller margin. **Turns to camera and speaks in Kevin Spacey voice** “Now if y’all’r expectin’ a House of Cards reference here, you had best look elsewhere.” Cersei Lannister is the far more interesting ruthless, conniving climber on prestige television right now. You get all the single-minded ambition plus frickin dragons. I’ll take my dystopian political dramas sword and sorcery style, thank you.

 

I promised some pretty graphs, and I will deliver at least on the graph part. I’ll leave the art criticism to you. First, a histogram of what the model thinks the national environment looks like in 2014, 2016, and based on these three special elections in 2017.

samphist

Next, a plot of quantiles of sampled elections given those sampled national environments. You can read  this as basically what my model thinks the hypothetical 2017 general election looks like given these special election results. Each vertical line of 5 dots is one national environment sample. Each dot is the .05, .25, .5, .75, and .95 quantiles of election samples within that national environment.

sampled2017

Insert standard disclaimer here. November 2018 is not June 2017. I have no idea what politics will look like in a year and a half. This is an analysis of a purely hypothetical election happening now. Also, for the record, none of this should be particularly surprising given other indicators of current politics. Donald Trump’s approval rating is abysmal. The generic Congressional ballot in polls shows a hefty Dem lead. Talk of an enthusiasm gap has started. I’ll leave legal comments to people who know more about the law, but it’s pretty hard to say that accusations of obstruction of justice and collusion with a foreign power are good for a party. Nor is support for wildly unpopular legislation dealing with healthcare. Plenty can happen between now and the election, but right now, Republican electoral prospects do not look great.

 

 

*One thing that goes into the error term here, that probably shouldn’t, is incumbency advantage. Take all my results here with a grain of salt, because my model considers all seats to be competitive on equal ground, as if they were open seats. In reality, Democrats might have to do even better nationally to have the sorts of gains predicted here, because the relevant seats they have to gain are currently held by Republican incumbents.

**The method I use here is a Markov-Chain Monte Carlo simulation implemented using the JAGS package in R, which sounds way more impressive than it is. Seriously, this is a super simple model that started as a class project. I just thought the results were interesting enough to post about.

Advertisements

Trump’s Approval Rating 4 Months In

Yeah, it’s only been 4 months. Really it has. Four months and a week if you want to get technical.

President Trump’s approval rating has never been very good, but I wanted to look at how it’s changed over the past few months and who the people are that have moved against Trump.

Trump_JApproval1

Those outliers on approval come from a pollster called iCITIZEN who I’ve never heard of, but Huffpost Pollster believes is good enough to put in their aggregator so I kept them. Trump’s approval rating has fallen about 5 points from ~47% to ~42% give or take. His disapproval has risen from ~47% to ~55%. A net negative approval shift of about 13 points. It’s not good by any stretch of the imagination, but it isn’t as bad as it ought to be.

Given how polarized the parties are, the movement itself isn’t coming from Republicans or Dems, both are essentially unchanged. Let’s take a look at the independents. If the approval of Trump from Independents has fallen, that should be a good test of actual declines in his support. Rising disapproval from Independents shouldn’t tell us that much. Most of that movement is coming from undecided Independents who aren’t likely to participate in elections, particularly midterms. So if the politically disinterested independents dislike Trump, we shouldn’t gleam too much from that. We need to see declining approval from independents to start to see a real erosion in Trump’s support, given that Republicans don’t seem to keen on turning on him anytime soon.

IND_Trump_App

Again, if you ignore the outliers, there isn’t much movement in Trump’s Approval with Independents. It’s early, and there’s certainly time for his support to wane, but it hasn’t really happened yet. This raises questions about what parts of the news cycle are filtering down to the less politically engaged. Is it as negative as it should be? Are Trump’s Executive Orders, which are just glorified press releases getting positive local coverage? Can a politically disconnected person name two things about the Trump-Russia scandal? Does Trump’s twitter feed get his talking points out in an unencumbered way that presidents haven’t been able to do before? IDK. But if the economy continues at its current pace, I’m not sure Trump’s approval will move much.

IND_Trump_App1

Sans outliers. The scale gets changed here, which shows more movement than there actually is. Loess functions get weird when there are few observations at the end of the time series. Trump’s approval isn’t great, though it never has been. When is the bottom going to fall out? Will it? I’m concerned.

Using Special Elections to Analyze the National Political Environment

Guest Author: Danny Metcalf

 

As anyone who is reading this already knows, there’s a special election to fill the open seat in Montana’s at-large House district today. It is the third high profile House special election this year, the first two being KS-04 and GA-06 (the Georgia race has a runoff on June 20th). While the seats themselves matter, most of the focus on these races has been trying to divine something about the national environment. We want to see John Ossoff do well in Georgia’s 6th and say that Democrats across the country will do well. Unfortunately, a lot of the analysis on these races has been poorly thought out.

 

The worst commentary is the, “Sure Democrats are doing well, but they’re not actually winning. That shows Republicans can hold on, despite their president’s struggles,” type. Slightly better commentary recognizes the obvious problems with the first type and then tries to just assume a uniform national shift: Dems did roughly 20 points better in KS-04 than they did in 2016, here’s what the map would look like if they did 20 points better across the country. Unfortunately, that makes another sort of obvious error; it ignores regression to the mean. Imagine you’re waiting for results to start coming in tonight, so you turn on the Celtics-Cavaliers basketball game. In the first quarter Lebron James puts up 20 points (a big number for one quarter, but not out of the question for him). You should not look at that boxscore and think “Lebron’s getting 80 tonight.” Okay, bad example, because Lebron is a superhero and he could totally put up 80, but hopefully you get the point. If a result surprises you it means it is higher or lower than you expect. Assuming your expectations weren’t just pulled out of thin air, you should expect future results to be somewhere between your original expectations and the surprising result. If Lebron gets 20 in the first quarter, he’s probably having a better game than his average, but he’s not scoring 20 every quarter. Similarly, if Dems do 20 points better than expectations in KS-04, they are probably doing pretty well nationally, but it’s crazy to expect them to outperform their baseline by 20 points in every district.

 

I’m going to present the results from a sort of analysis that tries to get around this problem. Bayesian statistics takes seriously our prior understanding of how random variables are distributed. I use computing tools to define a model for how House election results are generated, sample 1000 potential values for the national environment given the first two special elections, and then use those values to sample 100 election results for each House district. The results should give me a fairly strong look at what an election that produces these special election results looks like.

 

The first takeaway is that there is a lot of uncertainty. Even with two undeniably good results for Democrats, we still have very little data. These are the sorts of results you would expect to see in a wave election, but they are also consistent with moderate gains for Democrats, and they could even be outliers in a slightly Republican year. The 95% credible interval (analogous to a confidence interval in frequentist statistics) is from 189 Democratic seats (a small decrease for Dems) to 274 seats (a landslide victory). That said, there are some things we can say about the type of electorate that produces these elections. The median number of Democratic seats won is 236 seats, substantially more than the 218 needed to claim the House. In fact, 76.8% of sampled elections end up with at least 218 seats for the Democrats.

 

Those numbers are obviously encouraging if you’re a democratic operative, but I do want to sound a note of caution before overreacting. The way my model works is it considers a potential national environment that produced these results and predict what the rest of the country does in the same environment. There is absolutely no reason to believe, and there are any number of reasons to doubt, that the 2018 general election will take place in the same sort of national environment as these special elections. 18 months is a lifetime in normal politics. At the risk of editorializing too much, a single week of Trump time feels like years of normal politics. I make no effort in this analysis to predict the future. I am only trying to get a feel for what the national environment looks like now.

 

The other reason to be cautious is the chance that these special elections are systematically different from a normal House election. The occur roughly one at a time and therefore garner outsize attention. It is at least worth considering that they work in a different way from how normal general elections work. Then of course there are the added complications of GA’s group primary system and the allegations against Gianforte in Montana. While these are all potential concerns, I would actually advise assuming there is no major difference over assuming unique events. It is often easy to come up with reasons why any particular instance is different, but it is much harder to predict what particular variables actually matter to interpretation. I am comfortable subsuming all these concerns in the error term of the model and assuming that error size does not differ substantially between special and general elections.

 

*All of this analysis is predicated only on the two races for which we have results so far. I will redo the analysis once the Montana race is in and post an update, hopefully with some pretty charts and graphs, here.

*For a more detailed breakdown of methodology and code see Special Election Bayesian Estimates

Obama and Paid Speeches

Since this has become a serious point of debate this week I thought I’d address it even though there are real things going on like the government only being funded through next Friday, possible war on the Korean Peninsula, a Tax plan so poorly fleshed out Sean Spicer thought tax preferences for 401Ks had been eliminated, and another attempt to take away health insurance from low-income people with the even more draconian AHCA II: The Deathening.

For those who got lost in the impossible to keep up with news cycle, President Obama is said to give a paid speech to Cantor Fitzgerald for a whopping sum of $400,000 in September. Obama, who will likely never hold elective office again unless the South secedes for a second time, is an interesting person to try and corrupt. If you believe Bill and Hillary were compromised by paid Wall St speeches, it is a bit curious to worry about President Obama given that his wife has NO INTEREST in running for office and his kids are 16+ years away from even being eligible to run for President and are likely smart enough to know that happiness exists outside of public life at the highest levels. The Clintons do not know this.

From a quid pro quo corruption standpoint there is absolutely nothing wrong with President Obama giving paid speeches given his lack of formal power to you know, make policy or change laws. He will always have informal lobbying power, but his 3 month long vacation suggests to me that he’s not super interested in wielding it. Matt Yglesias, however, makes a really strong case against Obama making paid speeches. The crux of his argument is not that Obama was or is corrupt but rather making paid speeches provides fodder for many on the left who see the technocratic center-left as beholden to the interests of their donors, and thus sells out the working class base to keep financial and political elites happy, i.e. Obama, Hillary, Schumer, Pelosi, etc. are all ‘on the take’.

From an empirical standpoint I certainly think this is false. Not just that they’re not being paid on the side to write ‘awful’ policy, but that to the extent that they have policy differences with the base, it’s driven by ideological distinctions rather than the machinations of the donor class. Obama is a moderate pragmatist who governed like a moderate pragmatist. It’s more satisfying to say he was corrupted by elites, but it’s more accurate to note that Obama is not an ideological firebrand and he never was. In so many ways Hillary who may be more ideologically conservative than Obama in areas like foreign policy, is in many ways a much more radical candidate who through paid family leave and support for equal pay, upends the social order in ways Obama doesn’t. (I’ll note that not many people agree with me on this point).

However, there’s a perception argument here that Yglesias and others point out that is absolutely worth addressing. I don’t have any problem with President Obama giving paid speeches. BUT I do believe that giving paid speeches will limit his political influence. Even though Obama was a president with virtually no ethical or moral scandals of any kind, and was a model citizen, he loses that goodwill while giving speeches, particularly to wealthy donors. IT WILL degrade his influence both with the left and with Congress, who increasingly faces a populist backlash from voters.

The biggest mistake I made as a political strategist and amateur pundit in 2016 is I explained away the data in front of me. I got to see some internal HRC polls in July of 2016. THEY WERE HORRIBLE! If I had seen these polls for any other candidate for any other race in the country I would have said they were doomed. I never seriously considered that other people didn’t like her. I love Hillary Clinton. I really see here as an ideal President even if she’s not an ideal candidate. I never truly entertained the idea that people wouldn’t like her despite mounting evidence that the public hated her.

Obama built an enormous amount of goodwill with what will likely be considered a massive success of a presidency. He can’t afford to corrode that by giving paid speeches and pocketing the money. If Obama gives paid speeches and donates that money to charity, that’s totally fine. If people see him as on the take, he’ll have a profitable but empty retirement. I don’t think that’s fair, but I do believe it’s the standard he’ll be held to.

 

 

Dems Shouldn’t Spend Money in KS-04 MT-01 or GA-06

This will be the hottest establishment take you read this week, but spending money in these places is dumb.

Let’s take a step back from these races. Why do we have the DNC, and it’s Congressional campaign arms the DCCC and the DSCC? For Democrats to control Congress. That’s why these institutions exist. They fundraise, recruit candidates, and target voters to win elections. That’s their job and it’s why they were created. If your goal is to control the House, so 218 seats, how do you do it? You control every district that has more Democrats than Republicans, you win as many of the tossup districts as you can, and you win the winnable Republican leaning districts, those with R+1-R+5 PVI scores. The Cook Political Report’s Partisan Voter Index (PVI) is the most useful tool for understanding this concept. They have paywalled the 115th Congress scores, but it’s easy enough to calculate on your own. Here’s the link the ranked districts by PVI score. In the 113th Congress, where Republicans had a 234-201 margin, Republicans controlled only 4 Democratic leaning districts, whereas Dems controlled 8 Republican leaning districts. Even Districts were split 6-3 Dems.

You might be wondering, how is it possible for Dems to hold more Rep seats than Reps holding Dem seats, but Gerrymandering is a fascinating thing. Republicans have drawn lines so well, they basically have 230 districts that Dems aren’t penetrating without a wave election meaning a big Democratic swing. But there are 12 R+1 districts, 14 R+2 Districts, 9 R+3 Districts, and 12 R+ 4 districts. Of these 47 Districts, Dems only held 6, leaving 41 winnable districts for Dems to target, throw in the other 3 Even districts and we have 50 seats to target. In 2014, Dems only needed to win 18 more of these to take back the House (they lost 7 seats that election cycle). These are the ‘winnable’ districts where Dems and the DCCC should spend every conceivable dollar.

Ossof’s district, Georgia’s 6th, is a PVI of R+8. R+8 in 2014 would have been the 87th Most Winnable Republican leaning seat in the country. If we’re winning GA’s 6th, we’re winning the overwhelming majority of the country.

I fundamentally do not believe there is anything to be gained in fighting for moral victories that will be corrected in 18 months. It might make you feel good. It might be fun to fuck with Republicans in a R+15 district. But it isn’t what a responsible strategic party organization does. The goal is to control Congress, not chase shiny objects when they start to appear. If you want to criticize how they raise or spend money I’m all ears, but targeting KS-04, MT-AL and even GA-06 is just silly. it’s acting for the sake of acting. It’s not motivated by theory or goals, it’s motivated by emotion, which is how you make stupid, costly errors, like a primary challenge to Joe Manchin.

I’m willing to entertain an argument that the Democratic Party as an institution needs to completely rethink how it engages with voters. That a successful future Democratic Party has to develop a language and a presence in more rural, largely white enclaves that aren’t highly educated if they want to win national elections. I’m even willing to entertain the idea that the Democratic Party as a whole needs to emphasize class politics and de-emphasize identity. But there just isn’t a viable strategy for Dems to compete in a district where Trump got over 70% of the vote beyond a one-off special election where turnout is artificially low.

Policy in a Polarized Era

In between New York Times push notifications about apocalyptic events, there actually is some policy-making going in Congress. By policy I mean House Republicans’ decision to either repeal the Affordable Care Act (ACA) in its entirety, or maintain most of the law but make a handful of fringe changes that just make health care marginally more expensive for poor people.

The figure below, approval of the ACA, I think gets at the two major concepts that underlie public opinion polling. The partisans are highly stable in their attitudes, following cues from party elites on what to support, and the ‘persuadables’, however they’re defined, are unstable because they don’t know what elite cues to follow and their reference points (could be thought of as primes) change their answer.

 

aca

Nothing changed or came out about the ACA that should have changed attitudes toward the law in January of 2017. Donald Trump became president and favorability of the law improved ~8 points. If you’re a ‘persuadable’ how does President Trump change your attitudes toward the ACA? For starters, the reference point for people who are not diehard partisans has changed. In 2014 if you’re asked about the ACA most stories focus on premium costs for people who purchased insurance on the exchange. The information that was primed revolved around premiums and what those premiums covered. There was much less focus on the Medicaid expansion at that time, or the people who gained insurance who didn’t have it before. Premiums, which are expensive and no one enjoys paying prime negative attitudes toward the law, even though for many families paying expensive premiums is a million times better than having no health insurance at all.

In 2017, the relevant media frames are how will Republicans insure the ~20 million people who would lose insurance by removing the exchanges and ending the Medicaid expansion. The question wording is quite literally the same. The polling firms are using the same methods. The law hasn’t been updated. It’s one of the most covered policy issues of the last 8 years. Attitudes shouldn’t move much without new information and yet, what’s shifting public opinion on the margins is likely the point of reference from media coverage. Premium costs v. losing health insurance altogether. We have a new frame, and attitudes seem to be reflecting that.

It’s almost a natural experiment in media coverage. With the presidency exchanging parties, repealing the law became a real possibility and thus the focus of the coverage of the ACA changed. Attitudes of the movable people went with it. As someone who tends to downplay media effects, because you have to see/read the news in order to be persuaded by it, it seems clear that shifting frames, particularly on highly covered policy issues, can move attitudes just because they change the public discourse around the issue. Maybe media effects are real, or rather they’re real for a tiny, but extraordinarily important segment of the public. Changing the reference point might be the key to influencing public opinion in a polarized era.

We Just Instituted a Ban on Muslims

Through an executive order, President Trump just banned entry to the United States for all refugees, and Muslims from 7 countries. On the one hand, this shouldn’t come as a surprise given that Trump explicitly campaigned on doing so for all of 2016. However, Republican White Nationalist rhetoric in the post 9/11 era has been mostly bark, with little policy bite. Yes we created the Department of Homeland Security, and yes the US needlessly invaded Iraq, but Republican elites have for the most part built up white nationalist sentiments, without ever following through on that rhetoric.

Not only did the muslim ban poll pretty well back in June (50-46 in favor), better than the ACA does today, but for a large segment of the public Islamophobic sentiments have stewed for quite some time. In polls or focus groups, people will express deplorable sentiments toward Muslims, but often that rhetoric doesn’t have any policy content behind it, just the kind of casual racism that pervades a good chunk of the white public in America. Hell, in a campaign I worked on, I ran into Sarah Palin Republicans and Bill Maher Democrats who had eerily similar things to say about Muslims.

To summarize: A lot of Americans (about half, maybe a bit more) dislike Muslims. ➡️ Republicans have tapped into this sentiment by highlighting “Radical Islamic Terrorism” in order to fundraise and mobilize voters. ➡️ Republicans in office, pass a few resolutions, but mostly try to cut taxes and spending, while paying lip service to national defense and other images strongly associated with white nationalism. ➡️ Rinse and repeat in the next election cycle. This can work for a while, but it’s certainly not a process that can sustain itself.

Republican strategists seemed comfortable operating in his pattern, concluding that they could ride this anti-Muslim sentiment (not to mention racism towards all other non-white groups) to electoral success, but never have to go as far as supporting a Muslim ban, or deporting/rounding up Muslims. The GOP knew what they were doing, but they still felt like they could stay in control of it. The 2016 Presidential Primary proved that theory false. This Mike Pence tweet embodies the idea that they were in control.

I want to make a caveat that this isn’t entirely the fault of Republicans. Americans don’t need a political party to help them come to racist conclusions, people are good enough at doing that on their own. However, Republicans did not do enough to try and stamp out the Islamophobic fervor that has animated their party since 2001, and many GOP leaders stoked it for national prominence. It was certainly going to take a shameless, racist, power-hungry, imbecile to capitalize on this opportunity, but even ignoring Trump, it wasn’t as if Governor Huckabee, Ben Carson, Newt Gingrich, or Rick Santorum weren’t dabbling in the same racist pools as Trump. Republicans helped create this monster, not the Trump monster, I mean the seething white-nationalist masses that are comfortable with deporting/banning people based on race, religion, or national origin. This sentiment was always present, as the paranoid style of American politics always is. But Frankenstein lost control of his monster, and he’s seemingly washing his hands of any responsibility over it. Conspicuously silent Speaker Ryan is seemingly willing to trade tax cuts for people’s lives. I just hope he’s knows that’s what he’s doing.

So, uh, What Happened in Wisconsin?

Let’s start out with a thought experiment. Let’s pretend for a moment that up until 2016, Wisconsin was a US territory that had just been admitted into the union by a bipartisan coalition of legislators. The territory of Wisconsin had elections, but their party system was different than ours, so we only knew about Wisconsin’s demographics, like race, education, income, and religion(s). If we had no prior electoral information about the state, we’d likely think Wisconsin would be competitive with two major cities supporting Democrats (Milwaukee and Madison), a large wealthy white suburb supporting Republicans (Waukesha) and a hodgepodge of support in the rural, predominantly white parts of the state. If we took Milwaukee, Madison, and Waukesha as fixed entities, we’d conclude that if the rural parts of Wisconsin voted more like it’s neighbors Iowa and Minnesota, that the state would lean Democratic, but be competitive with the right Republican candidate (i.e. not Romney). Through this frame it’s understandable how Trump, or the right Republican could win Wisconsin, even if past electoral history biased us against it. A caveat here is that Bush almost won Wisconsin in 2000, he lost by only 5,000 votes out of 2.5 million, 0.22% about the same percentage as Trump’s margin in Michigan. What happened in Wisconsin three weeks ago seemed impossible, but it’s not as crazy as it seems.

By now, you’ve read 25 stories on how working class white voters (loosely defined?, undefined?) broke for Trump in the midwest because of racism, sexism, xenophobia, economic anxiety, the opioid epidemic, desire for change, complete lack of faith in the political system, hatred of banks, etc. I’m not going to attribute any one solely to Trump’s victory, but they all certainly played some role. There’s even a lot of talk about Obama-Trump voters, a proposition that I want to rule out a priori because of how absurd it seems, but that we won’t be able to confirm until after the voterfiles are updated. If there’s a sharp increase in turnout of people who didn’t vote in 2012, and a decrease in turnout among people who did vote in 2012, what looks like Obama-Trump voters, might just be differential turnout. Of course some people voted for Obama and then Trump, but I want to be careful before assuming that’s true and what drove the difference in Wisconsin. In Milwaukee Clinton got 43,000 fewer votes than Obama, but Trump got 26,000 fewer than Romney. There likely wasn’t much vote switching, but depressed turnout explains a lot in Milwaukee, and might explain behavior in other parts of the state.

On paper there’s virtually nothing different about Wisconsin in 2012 than in 2016. Scott Walker is still Governor, the state’s unemployment rate is lower, but still behind the national average, Republicans still control the state legislature, and state GDP growth stayed at it’s 1% level almost the entire time. There certainly weren’t any national or state shocks that should have altered voting in any meaningful way. But one relationship pops up that’s fascinating: The relationship between education and the Republican candidate’s share of the vote.

Below are scatter plots of Romney and Trump’s county level Two Party Vote Share ( Dem Vote / (Dem Vote + Rep V0te)) labeled TPVS by plotted against county-level College Attainment (percent) and Median Individual Income as measured by the Census. Two Party Vote Share is the x axis for all of these plots below.

romney_income

trump_income

romney_education

Trump_Education.jpg

For those of you that have read Andrew Gelman’s Red State Blue State Rich State Poor State these figures should be familiar. Romney’s vote share and income relationship, matches the relationship we’d expect in a solidly Republican state like Texas or South Carolina, as voters get wealthier they become more Republican. But Trump’s relationship with vote share and income looks like that of a solidly Blue state like Connecticut or California, as income increases, there’s no relationship with partisan voting. Remember, all 4 of these graphs are the same 72 counties in Wisconsin from 2012-2016. And they’re of the same fucking party.

To recap, county-level median income is correlated with Support for Romney, but not for Trump. County-level College attainment is negatively correlated with support for Trump, but not for Romney. An extraordinarily important caveat here is that this is explaining how counties voted, not how individuals behaved. So Trump does worse than Romney in Waukesha, a well-educated affluent suburb, but he still gets a ton of votes from there that gets downplayed when you look at counties instead of individuals. Below is a link to a downloadable map to look at this data for 2016 graphically. You can download it and view it in Chrome.

Map: 

The simplest interpretation of these data is that in 2012 economic concerns heavily influenced voting behavior in ways that fit the traditional partisan narrative, Republicans are wealthier and Democrats are poorer. 2016 shows that voting behavior was driven most heavily by differences in educational attainment (a proxy for social and cultural issues or cosmopolitanism, rather than literal education), with income being a relatively unimportant determinant of voting behavior, a kind of bizarro 2000.

As a baseline test I thought I’d regress Trump Vote share on county level Education and income, to see what patterns emerged. In the url for the .PDF below are a series of weak, but I think instructive models looking at the relationship between income and education on vote share. The big thing that seems to stand out is that income, as measured at the county level, is not driving support for Trump. I wouldn’t read this as “there’s no economic anxiety, they’re all racists” but the story of economic anxiety should not be considered the sole driver of voting behavior in the midwest, as at least some pundits would like us to believe.

wisconsinmodels

I don’t know how we define class in the US, and I’m increasingly unsure what others mean by it when they use it. It’s sort of become this stand-in for education, but there’s a real cultural component to it as well. It’s not just whether or not someone has a college degree, but watching football or CBS sitcoms puts someone in a different class than if they watch Game of Thrones and read Vox. What I’m getting at is there is a growing cultural divide that expressed itself throughout the Midwest this past election. It’s a lot more complicated than “economic anxiety” and it won’t solve itself simply by redistributing wealth, even though we obviously should. There’s a lot left to unpack from this election, but the notion that this is a purely economic phenomenon just does not seem right to me.

 

Dems in Disarray!

Apologies for the title, but I can never pass up an opportunity to use my favorite phrase in American Politics. In Democratic networks, the circular firing squad has already begun, with the chief dispute surrounding the race v class divide that was relatively dormant until the primaries started last Fall, one that I talked about in a post earlier this year. Hillary Clinton supporters like myself argued ad nauseam that the real divide in American Politics has and always will be about race, with a growing divide along the lines of gender, as the Christian right became fully captured by the Republican Party. It is virtually impossible to think about the construction of American Institutions without race, from slavery, to the end of reconstruction into the Jim Crow era, to voting rights, to red-lining, to housing and workplace discrimination, to the drug war, and the dissolution of the welfare state, these policies were driven either entirely by race, or by the southern Democratic Party’s maintenance of white supremacy.

That’s why for many the rise of Bernie Sanders was such a shock. Here we have this white populist running at the end of a popular two-term Democratic Presidency with a large economic expansion, harping on the rigged economy, and the need for a Social-Democratic Revolution™ in America. By no means was Bernie Sanders the first left-wing white populist in America. In North Carolina in the 1880s, white farmers, struggling from high tariffs and the recession, began arguing for price supports for food, and state aid to survive. The Robert LaFolette populist movement created the Wisconsin idea, the partisan primary, and the roots of the direct election of Senators, the 17th Amendment. But populism in the latter half of the 20th century was found mostly in the Republican Party, expressing itself as nationalism and bordering along jingoism during the first term of George W Bush’s administration. There has not been a tangible left-populist movement in America for quite some time. It’s one of the reasons why I’m willing to forgive the Clinton campaign for not running a populist message. On the one hand, no one’s going to buy Hillary Clinton the populist, she couldn’t be a worse messenger for that movement. But more importantly, at my core I’m very skeptical of an economic populist movement that would incorporate and represent people of color, and create winning electoral coalitions. It’s not that people of color wouldn’t benefit from a European-style welfare state, but that historical expansions of the welfare state intentionally left people of color out of the loop. African-Americans were excluded from the Homestead Act of 1862, The Federal Housing Authority refused to back mortgages sold to African-Americans, and the Social Security Administration excluded jobs traditionally held by African Americans. African Americans did not get to participate in The New Deal writ large. Universal college funding and debt-forgiveness for student loans would have big economic benefits for lots of Americans, but they’d also be a transfer of wealth from working class people to middle class college educated Americans, who in spite of their debt, need less help than their working class counterparts. A white-populist movement on the left has a lot of good ideas, but they’re certainly oriented away from people of color.

I’ll make my biases known. I’m not particularly enthusiastic about an America where the new right movement is driven by nationalists and white supremacists, and the new left animated by populists and social-democrats. I OBVIOUSLY prefer the new left, but it’s not my ideal coalition. I have to admit I liked the multicultural neoliberal movement we had going, but I recognize that it’s a political movement that’s dying in the western world, and may not be rebuilt in America without President Obama at the helm. I certainly don’t think Bernie Sanders is some closet racist who wants a whites only welfare state, but I don’t believe that it is a completely random turn of events that the places where Bernie Sanders has the most support are overwhelmingly white middle and upper-middle class enclaves (Portland, Seattle, Madison, Manchester NH, Minneapolis). My biggest concern with Bernie Sanders as the Democratic nominee is that for every working class white voter you might pick up in MI, WI, PA, IA (States we lost in the last election) you stand to lose college-educated whites in the suburbs everywhere, who also have higher probabilities of turning out to vote. The college-educated white voters who live in the suburbs, don’t do so randomly either. They flee cities to have quasi-private public schools, that through busing and school-zone districting are effectively racially segregated. Ironically, the social-democratic movement in the US that is being promoted largely by younger well educated white people (i.e. Bernie Bros™), directly appeals for a more advanced welfare-state, that scares-off older suburban whites, because of their latent racial animosity, fear of the “welfare-queen”, and belief in the low-tax fetishism sold to them by the GOP. I know that last sentence sounds ridiculous, but if exit polls are to be believed and Trump won college-educated whites by 6%, there are real concerns about selling any form of economic populism to a larger electorate. Trump doesn’t win working class whites because he promised them new manufacturing jobs he won’t deliver on, or a beautiful wall on the Mexican border he won’t actually build, he wins them because he offered them cultural primacy. He told them Elites were out of touch, and that “the real hardworking good people in America” live in small towns, forgotten by coastal and academic elites (not all that different than Bush’s reelection campaign in 2004). He offered them respect (which by some is read as white supremacy) and that’s something that Democratic elites certainly do not convey (of this I am extremely guilty).

The last year has been a humbling experience for me. Up until the very end, I thought Hillary would win handily, and I certainly was duped by the “Demographics is Destiny” line of thinking in the party. I will no longer speak with the certainty and arrogance I did over the past year, but I’m not going to stop writing either. I’m still skeptical that Keith Ellison will be able to pull off the multiracial populist coalition at the DNC that he’s successfully built in Minnesota, but I’m eager for him to lead the DNC and I think he’s a good spokesperson for the party. Hopefully, I’m wrong and we’ll be able to move left on economic policy, and win back governor’s seats in 2018, and the presidency in 2020. Let’s just say I’m not convinced as of yet.

 

Presidential Predictions

At a later date I’ll explain my 6 month absence from the blog, but I wanted to post my electoral college predictions for tomorrow. I want Hillary Clinton to win with every fiber of my being, but I think it’s going to be a nail biter. I see a 278-260 win for Hillary Clinton. Trump is going to gain Iowa, Ohio, Florida, and Maine’s 2nd Congressional District from Obama’s 2012 map.

My map is below. Leave your predictions in the comment section. Enjoy election day tomorrow!


Click the map to create your own at 270toWin.com