Two political science-y comments on impeachment
Fine. I'll do it. Comments on the news of the week.
First, let's address the challenge of polling on impeachment. Short version: be leery of it. I'll begin with a link to an article I wrote for The Conversation on polling... before the Ukraine phone call leaked (delay in publication). The Conversation asked me to write something very general on polling, how to analyze and interpret polls, and so forth, and that's what I gave them. It has some important lessons for polling on impeachment.
Specifically, asking voters about impeachment directly is, at best, a weak indicator. What's actually happening is that voters are lining up on impeachment, pretty much with their parties! That was my point! And... polling data can't tell us how they'd react after a failed impeachment (I'll get to that in a moment), partially because they don't necessarily know!
There are 13 months between now and the 2020 general election. Current polling tells us little about how a failed impeachment would play out, other than that voters are beginning to line up along party lines, more consistently than they did with the possibility of impeachment from the Mueller/Russia investigation, where party leaders didn't signal impeachment on the Democratic side.
Be careful with polling data here. The guidelines are in my Conversation article.
Next point: probability of acquittal. Let's talk, for a moment, about the Senate. I'm normally a pretty big fan of prediction markets. That's not to say they are infallible. After all, in 2016, they mostly just followed polling aggregators, and hence were just as error-prone as polling aggregation, when what they should have done was follow political science forecasting models, like Alan Abramowitz's "Time For A Change" model. Still, they can be useful. Prediction markets can be a useful baseline. Right now, though, they don't make a lot of sense, nor does a lot of the talk about the supposed possibility of Trump's conviction in the Senate.
Will Trump be impeached? I don't know. What if he were, though? Here is the PredictIt market for Trump completing his first term. Basically, 2-1 he scrapes through. Now, there are multiple ways to not finish a term. Put bluntly, there's the William Henry Harrison method, the James Garfield method, the Richard Nixon method, and the never before seen option about which people are now speculating-- impeachment+conviction=Trump? Kinda like R+L=J, but different. What do they have in common? They're both totally fictional! Zing!
Anywho, if you think that Trump will be convicted by the Senate, then riddle me this, Batman: which specific Republicans vote to convict? Here's my political science-y point for the day on this: unitary actor.
Sometimes, we reduce a group of people to a "unitary actor," for the sake of conceptual or mathematical convenience. In international relations, we will often do this with a whole country! But... a country is usually not one person. Unless you declare yourself a sovereign nation as some sort of tax scam, in which case, please consult your attorney, not me. I'm just a political scientist. In some bargaining models, we'll reduce "Congress" to a unitary actor as it bargains with the president. Spatially, we can kinda sometimes reduce Congress either to its floor median, or majority party median, or something, and pretend that we aren't doing horrendous damage to reality, but... we are. We just need to simplify the math. People like me have a bad habit of sacrificing realism for mathematical necessity. We suck.
The electorate? Funny thing, but we can almost, kinda think of the electorate as something conceptually like a unitary actor if we take a prominent model from MacKuen, Erikson & Stimson's Macro Polity, which essentially argues that the public collectively has moods for left or right policies that get satisfied, leaving the public to shift back and forth. (Ken Arrow has a little something to say about this, but I'm getting off-track here...)
But... a group isn't an individual, and there are lots of mathematical problems that ensue when you try to pretend that a group is an individual. And when you dig down at the level of 100, and when your goal is vote-counting... making a unitary actor assumption is just junk science.
And here's what I mean. A lot of the arguments I see floating around for why the Senate might convict Trump boil down to something about how much electoral pressure there may be on the Republican Party via polls (see above), weight of evidence, conflicts between Trump and the GOP that have been suppressed for years floating to the surface-- I'd argue Syria was the dumbest political decision Trump has made since Flynn-- and that kind of stuff.
All of this treats the GOP as one actor, where enough pressure on that one actor weakens support enough from that one actor that it ceases to provide enough support to maintain Trump's position.
But... that's not what the Senate is. What is the Senate? As the old joke goes, 100 people who all think they should be president. Most importantly, though... 100 people. Not 1... 100. That's a two order of magnitude difference.
So, which Republican senators, specifically, break off and vote to convict Trump? Let's break this down. First, suppose he actually were taken down. Pence would lose in 2020, 100%. Warren, then, (or whoever, but probably Warren at this point) faces a midterm in 2022 when the Republicans would have a midterm advantage. The only Republicans who have any possible electoral pressure to convict Trump are the ones up in 2020.
So... who? Mitch McConnell? Lindsey Graham? Joni Ernst? Tom Cotton? Jim Inhofe? Who? Gimme the names! Name names! Vote-counting is about naming names. Otherwise, we are committing a unitary actor fallacy.
Susan Collins is up, and there would be pressure on her, particularly with the fallout of her Kavanaugh vote, but a) that vote itself, along with her vote on the Tax Cuts & Jobs Act, which negated her vote against "Obamacare repeal,"* should remind Democrats that, at the end of the day, she's not with them, and b) as difficult as it is to get Collins, Collins alone gets them nowhere.
For the Senate Republicans, this isn't about whether or not to hand the Oval Office to Pence. It's about whether or not to hand the Oval Office to-- highest likelihood right now-- Elizabeth Warren, since convicting Trump dooms the party.
We need to remember what polarization is. The parties are very, very, very far apart in policy terms. Even if some of the Senate Republicans decided that they disapproved of Trump, the cost of removing him would be tremendous, in policy terms, because of the gap between the parties. For Collins, the gap would be smaller because she isn't as reactionary as, say, Tom Cotton or James Inhofe, but when the alternative is Elizabeth Warren, who is a political radical, the policy cost is too high for the current mainstream of the party, and maybe even for Collins.
So, at an individual level, who would peel off and convict Trump? Who would pay that cost?
Without specific names, and explanations for how those specific senators get there, this is all nonsense.
*The Obamacare repeal bill on which Collins voted "no" was the "skinny repeal" bill, which only repealed the individual mandate, but Collins then voted "yes" on the tax cut bill, which included a repeal of that very mandate.
First, let's address the challenge of polling on impeachment. Short version: be leery of it. I'll begin with a link to an article I wrote for The Conversation on polling... before the Ukraine phone call leaked (delay in publication). The Conversation asked me to write something very general on polling, how to analyze and interpret polls, and so forth, and that's what I gave them. It has some important lessons for polling on impeachment.
Specifically, asking voters about impeachment directly is, at best, a weak indicator. What's actually happening is that voters are lining up on impeachment, pretty much with their parties! That was my point! And... polling data can't tell us how they'd react after a failed impeachment (I'll get to that in a moment), partially because they don't necessarily know!
There are 13 months between now and the 2020 general election. Current polling tells us little about how a failed impeachment would play out, other than that voters are beginning to line up along party lines, more consistently than they did with the possibility of impeachment from the Mueller/Russia investigation, where party leaders didn't signal impeachment on the Democratic side.
Be careful with polling data here. The guidelines are in my Conversation article.
Next point: probability of acquittal. Let's talk, for a moment, about the Senate. I'm normally a pretty big fan of prediction markets. That's not to say they are infallible. After all, in 2016, they mostly just followed polling aggregators, and hence were just as error-prone as polling aggregation, when what they should have done was follow political science forecasting models, like Alan Abramowitz's "Time For A Change" model. Still, they can be useful. Prediction markets can be a useful baseline. Right now, though, they don't make a lot of sense, nor does a lot of the talk about the supposed possibility of Trump's conviction in the Senate.
Will Trump be impeached? I don't know. What if he were, though? Here is the PredictIt market for Trump completing his first term. Basically, 2-1 he scrapes through. Now, there are multiple ways to not finish a term. Put bluntly, there's the William Henry Harrison method, the James Garfield method, the Richard Nixon method, and the never before seen option about which people are now speculating-- impeachment+conviction=Trump? Kinda like R+L=J, but different. What do they have in common? They're both totally fictional! Zing!
Anywho, if you think that Trump will be convicted by the Senate, then riddle me this, Batman: which specific Republicans vote to convict? Here's my political science-y point for the day on this: unitary actor.
Sometimes, we reduce a group of people to a "unitary actor," for the sake of conceptual or mathematical convenience. In international relations, we will often do this with a whole country! But... a country is usually not one person. Unless you declare yourself a sovereign nation as some sort of tax scam, in which case, please consult your attorney, not me. I'm just a political scientist. In some bargaining models, we'll reduce "Congress" to a unitary actor as it bargains with the president. Spatially, we can kinda sometimes reduce Congress either to its floor median, or majority party median, or something, and pretend that we aren't doing horrendous damage to reality, but... we are. We just need to simplify the math. People like me have a bad habit of sacrificing realism for mathematical necessity. We suck.
The electorate? Funny thing, but we can almost, kinda think of the electorate as something conceptually like a unitary actor if we take a prominent model from MacKuen, Erikson & Stimson's Macro Polity, which essentially argues that the public collectively has moods for left or right policies that get satisfied, leaving the public to shift back and forth. (Ken Arrow has a little something to say about this, but I'm getting off-track here...)
But... a group isn't an individual, and there are lots of mathematical problems that ensue when you try to pretend that a group is an individual. And when you dig down at the level of 100, and when your goal is vote-counting... making a unitary actor assumption is just junk science.
And here's what I mean. A lot of the arguments I see floating around for why the Senate might convict Trump boil down to something about how much electoral pressure there may be on the Republican Party via polls (see above), weight of evidence, conflicts between Trump and the GOP that have been suppressed for years floating to the surface-- I'd argue Syria was the dumbest political decision Trump has made since Flynn-- and that kind of stuff.
All of this treats the GOP as one actor, where enough pressure on that one actor weakens support enough from that one actor that it ceases to provide enough support to maintain Trump's position.
But... that's not what the Senate is. What is the Senate? As the old joke goes, 100 people who all think they should be president. Most importantly, though... 100 people. Not 1... 100. That's a two order of magnitude difference.
So, which Republican senators, specifically, break off and vote to convict Trump? Let's break this down. First, suppose he actually were taken down. Pence would lose in 2020, 100%. Warren, then, (or whoever, but probably Warren at this point) faces a midterm in 2022 when the Republicans would have a midterm advantage. The only Republicans who have any possible electoral pressure to convict Trump are the ones up in 2020.
So... who? Mitch McConnell? Lindsey Graham? Joni Ernst? Tom Cotton? Jim Inhofe? Who? Gimme the names! Name names! Vote-counting is about naming names. Otherwise, we are committing a unitary actor fallacy.
Susan Collins is up, and there would be pressure on her, particularly with the fallout of her Kavanaugh vote, but a) that vote itself, along with her vote on the Tax Cuts & Jobs Act, which negated her vote against "Obamacare repeal,"* should remind Democrats that, at the end of the day, she's not with them, and b) as difficult as it is to get Collins, Collins alone gets them nowhere.
For the Senate Republicans, this isn't about whether or not to hand the Oval Office to Pence. It's about whether or not to hand the Oval Office to-- highest likelihood right now-- Elizabeth Warren, since convicting Trump dooms the party.
We need to remember what polarization is. The parties are very, very, very far apart in policy terms. Even if some of the Senate Republicans decided that they disapproved of Trump, the cost of removing him would be tremendous, in policy terms, because of the gap between the parties. For Collins, the gap would be smaller because she isn't as reactionary as, say, Tom Cotton or James Inhofe, but when the alternative is Elizabeth Warren, who is a political radical, the policy cost is too high for the current mainstream of the party, and maybe even for Collins.
So, at an individual level, who would peel off and convict Trump? Who would pay that cost?
Without specific names, and explanations for how those specific senators get there, this is all nonsense.
*The Obamacare repeal bill on which Collins voted "no" was the "skinny repeal" bill, which only repealed the individual mandate, but Collins then voted "yes" on the tax cut bill, which included a repeal of that very mandate.
Comments
Post a Comment