Why Democrats Lose: The basics of game theory
Time for...
Oh, never mind. Let's just do this. Here's the thing. Democrats are not amongst the great strategic thinkers of our era. Did you watch the Lewandowski hearings? Are you following the ins-and-outs of Nadler's "I'm doing something, and it's vaguely impeachment-related" process? Corey Lewandowski walked into that hearing with a clear plan to stonewall, and blow off any Democratic questioning. He basically dared them to do something about it.
In fact, Rep. Tom McClintock (R-CA) has "double-dog dared" Democrats to just stop with Nadler's games, or call for a vote on articles of impeachment for Trump.
Democrats are doing neither. What's going on here?
Let's turn to one of my favorite books: The Strategy of Conflict, by Thomas Schelling.
First Schelling reference on the new blog! Woo-hoo!
Schelling's goal was to explain the nature of bargaining and the role of threats in bargaining. His initial motivation was to explore the concept of nuclear deterrence, which was actually kind of a mystery, in strategic terms. Here's what I mean. Introduce nuclear weapons, and even conventional warfare between nuclear armed powers suddenly is off the table because if you wind up in a war, it can go nuclear, and then we all turn into cheez-wiz, and while Philadelphians think that garbage goes on sandwiches, the rest of humanity knows that it is inedible toxic waste. We got nukes, the Soviets got nukes, and despite the fact that after WWII, we and the Soviets were enemies (hey, remember that?), we didn't even wind up in a conventional war. Why not? Risk of cheez-wiz-ification.
Here's the problem, for Schelling. Imagine two people, handcuffed together, standing on the edge of a sheer cliff. One says to the other, "gimme $20, or I jump," do you give the guy $20? Ummm, if he jumps, he dies. Yeah, he takes you with him, but... he dies. That's not a credible threat, unless he's coo-coo for coa-coa puffs.
So, how do nukes prevent conventional warfare? It doesn't make any sense. No matter how badly you are losing a conventional war, if you fire the first nuke, the end result is that everybody dies in a global thermonuclear war. That'd be... bad. Nobody should ever press the button first, so the existence of nukes shouldn't change anything about conventional war!
And yet it did. We never had a conventional war with the Soviets.
Here's why. Schelling got a Nobel for this. (Pay attention, Jerry.) What if, instead of standing at the edge of a sheer cliff, the handcuffed adversaries are standing at the top of a hill? The hill slopes down, and gets ever-steeper the further down the hill you go. Take a step, and you're fine. Two steps... gettin' a little steeper. A few more, gettin' steeper. The more steps you take, the bigger the risk that someone loses balance, and both people take a tumble. Each incremental step increases the probability of someone losing balance by some small amount, and it's just random chance whether or not that loss of balance happens. There were a few times during the Cold War when random chance almost got us all killed...
Now, is it credible for someone to say, "gimme $20 or I take a step?" Maybe! Or, perhaps make a bigger demand than $20... Depending on the stakes, the threat to take a step can be credible, even when both people are sane.
That's the concept. Once Schelling introduced this concept to the game theory world, a bunch of math ensued. Two games, in particular, became canonical. War of attrition, and brinksmanship. In brinksmanship, the structure is as follows. Two players. Player 1 decides whether to play or quit. If Player 1 quits, Player 2 gets something. If Player 1 plays, "Nature" decides whether or not to create a disaster, which hurts both players. "Nature" isn't a strategic actor. Nature just rolls dice, and Einstein can cringe all he wants.
Come to think of it, from now on, whenever I insert "Nature" into my games, I'm going to rename it, "Einstein," with a footnote about the legal paperwork it has submitted to change its name to "Heisenberg."
Anyway, if the disaster doesn't happen, Player 2 decides whether to play or quit. Then,Nature Einstein, with an increasing probability of disaster every time we go through this cycle. You can change the structure in a variety of ways, and sometimes we'll have Players 1 and 2 both make a decision before Nature Einstein rolls the bones. I've taught it both ways, and fundamentally, it doesn't change the math.
War of attrition? Similar. Players alternately decide whether to play or quit, with each decision to not quit imposing a direct cost.
Both games are repeated games, as opposed to a one-shot, simultaneous move game, like the prisoner's dilemma. The way we solve these games is with a technique called "backwards induction." The math is sometimes ugly, but the concept is simple. You start your logical process at the end of the game, and work your way back to the beginning. Consider brinksmanship. One player, for whatever reason, may have more tolerance for risk than the other. Why? We can build that into utility functions. It's just the nature of preferences. There are plenty of reasons, but for whatever reason, one player will be willing to step farther down that hill than the other. (Further? It's a metaphorical hill, so... But I am speaking of a hill. Farther. I'm going with farther here. Distance, movement... farther. I don't care that the hill is imaginary and metaphorical, I'm going with "farther.") What does that mean? Let's say Player 1 will step farther down the hill than Player 2. If it gets to the round where Player 2 would quit, then Player 2 has just run the risk of a disaster. Before that, Player 1 will definitely have played. If there wasn't a disaster before that. And so on, back through to the beginning of the "game tree." If Player 2 knows that Player 1 will definitely go farther down the hill, then the rational move for Player 2 is to quit as soon as it is Player 2's turn to make a decision. Otherwise, all Player 2 is doing is running risk after risk of disaster, and either facing actual disaster or quitting after facing pointless risk. War of attrition gets solved the same basic way. The one who will quit in the end should quit as soon as it is possible.
Backwards induction. This is at the core of modern game theory, and goes by the fancy-sounding term, "subgame perfect Nash equilibrium," when you work out all the math with a bunch of fancy equations 'n stuff. Sometimes, you need probabilities, 'n stuff, and our boy, Einstein will get uncomfortable, but hey. Try running an actual experiment sometime, Al! Oops! Did I say that?
I still love you, Al!
Anyway, what does this mean, for strategic interaction? Simply, this. How does it all end? Work your way backwards.
Impeachment. This is an absolute, sure-fire loser. The reason Tom McClintock double-dog dared Democrats to impeach Trump is as follows. Suppose Nadler brings articles of impeachment. Here are the possible sequences. It goes to the floor of the House, and they fail to get a majority of the House. Why? Not all Democrats are on board! Why not? Mainly, a lot of them think (correctly) that it would backfire in 2020. So, Nadler loses on the floor of the House. Or, the articles pass on the floor, and Trump is "impeached." Next? Trial in the Senate. Absolute, 100% chance of acquittal in the Senate. I'm not even going to play around with the statistician's bet-hedging here.
Trace that back. It means that the Democrats either have to pull back from the impeachment "proceedings," or lose. Nancy Pelosi gets this, and has been telling the party not to pursue impeachment because she's doing "backwards induction." A lot of the party, though, isn't. They're just looking at the next move. That's how you make the wrong move in game theory.
Let's turn, now, to Corey Lewandowski, and in fact, the whole parade of witnesses brought in to "testify" before congressional committees. Trump has told his people not to answer any questions. Can he do that? In practical terms, yes. Legally? Um... the boundaries of executive privilege are...
Anyway, what happened with Lewandowski was 100% predictable. He was instructed by Trump to stonewall, so that's what he did. Was he in contempt of Congress? A pretty strong argument can be made, and everything about what he did was predictable. In advance.
So, what could Nadler have done? Inherent contempt. Simply voting to hold Lewandowski in contempt wouldn't do anything. Wag your finger... whatever. If the goal is to ask the DoJ to intervene on Congress's behalf... good luck with that. However, there is this little thing called, "inherent contempt." Congress declares a non-cooperative witness to be in "inherent contempt" of Congress, has its Sergeant at Arms arrest the witness, and throw that witness in Congress's own, little cell. Yes, it is sort-of possible for Congress to have someone arrested rather than having the executive branch do it, when they declare inherent contempt.
But... does that sound kind of risky to you? Like, maybe we don't know how that would play out?
A little... brinksmanship-like?Nature Einstein would have to make a move there to determine how that plays out! If Jerry Nadler decided to declare that Lewandowski's behavior constituted inherent contempt, and asked the Sergeant at Arms to haul Lewandowski off to a cell in the basement of Congress... I don't know what would happen! Neither does anyone else!
But here's what I did know from the beginning. Nadler would never run the risk of trying. He would never step that far down the hill. Lewandowski, on the other hand, absolutely would stonewall. That made the outcome of Lewandowski humiliating the Democrats by stonewalling and getting away with it absolutely inevitable. Just as every other Trump official and former associate has succeeded in doing. Remember when Matt Whitaker started calling "time" on the Democrats questioning him? In every one of these hearings, we know what will happen. Trump's people will refuse to answer any questions. They will stonewall, and run down the clock. And, since the Democrats won't use inherent contempt due to their unwillingness to run the risk, that means the act of bringing in a witness like Lewandowski is pursuing a losing strategy.
Democrats aren't applying the logic of backwards induction. If they were, they wouldn't bother with any of these hearings. That doesn't mean hearings can't be useful, but if you know a witness won't cooperate, and that the DoJ and courts won't cooperate, then taking inherent contempt off the table means that hearings only matter if you have cooperative witnesses.
So, what do Republicans do? Here's the thing. When someone doesn't pursue a course of action because they've done the backwards induction, then by definition, you rarely notice. It's a dog that doesn't bark, much of the time. It isn't that the GOP just kind of rams through everything. Instead, consider the following.
2017-8. A Republican President, House and Senate. What didn't happen that Republicans would have wanted? What didn't they even try? How about privatization of Social Security? Just as an example. This has been floating around in the undercurrent of the conservative movement for decades, but every time it rises to the surface, the GOP gets slapped down hard. By backwards induction, what would have happened if they had tried in 2017 or 2018? Well, it would have failed because they couldn't have gotten a Senate majority, and then the unpopularity of the proposal would have hurt the party in 2018. So, they didn't do it. They didn't even try. That's backwards induction.
And because the party used backwards induction, unlike the Democrats, you didn't even notice. The dog didn't bark. Now, you may be thinking that the party still lost the House, and there are plenty of factors, but what would have happened if they had pushed on Social Security privatization? It would have been far worse for them. The value of practicing backwards induction is not facing that consequence.
On the other hand, the Democrats keep walking into losing situations by failing to trace out the obvious implications of their decisions. These hearings... we know how they go. Uncooperative witnesses will stonewall, and the hearings are pointless. The Democrats will not be able to remove Trump through impeachment. Trace this out. The math isn't that hard.
And read Thomas Schelling.
And for those still finding their way to the new blog, welcome, and spread it around.
Oh, never mind. Let's just do this. Here's the thing. Democrats are not amongst the great strategic thinkers of our era. Did you watch the Lewandowski hearings? Are you following the ins-and-outs of Nadler's "I'm doing something, and it's vaguely impeachment-related" process? Corey Lewandowski walked into that hearing with a clear plan to stonewall, and blow off any Democratic questioning. He basically dared them to do something about it.
In fact, Rep. Tom McClintock (R-CA) has "double-dog dared" Democrats to just stop with Nadler's games, or call for a vote on articles of impeachment for Trump.
Democrats are doing neither. What's going on here?
Let's turn to one of my favorite books: The Strategy of Conflict, by Thomas Schelling.
First Schelling reference on the new blog! Woo-hoo!
Schelling's goal was to explain the nature of bargaining and the role of threats in bargaining. His initial motivation was to explore the concept of nuclear deterrence, which was actually kind of a mystery, in strategic terms. Here's what I mean. Introduce nuclear weapons, and even conventional warfare between nuclear armed powers suddenly is off the table because if you wind up in a war, it can go nuclear, and then we all turn into cheez-wiz, and while Philadelphians think that garbage goes on sandwiches, the rest of humanity knows that it is inedible toxic waste. We got nukes, the Soviets got nukes, and despite the fact that after WWII, we and the Soviets were enemies (hey, remember that?), we didn't even wind up in a conventional war. Why not? Risk of cheez-wiz-ification.
Here's the problem, for Schelling. Imagine two people, handcuffed together, standing on the edge of a sheer cliff. One says to the other, "gimme $20, or I jump," do you give the guy $20? Ummm, if he jumps, he dies. Yeah, he takes you with him, but... he dies. That's not a credible threat, unless he's coo-coo for coa-coa puffs.
So, how do nukes prevent conventional warfare? It doesn't make any sense. No matter how badly you are losing a conventional war, if you fire the first nuke, the end result is that everybody dies in a global thermonuclear war. That'd be... bad. Nobody should ever press the button first, so the existence of nukes shouldn't change anything about conventional war!
And yet it did. We never had a conventional war with the Soviets.
Here's why. Schelling got a Nobel for this. (Pay attention, Jerry.) What if, instead of standing at the edge of a sheer cliff, the handcuffed adversaries are standing at the top of a hill? The hill slopes down, and gets ever-steeper the further down the hill you go. Take a step, and you're fine. Two steps... gettin' a little steeper. A few more, gettin' steeper. The more steps you take, the bigger the risk that someone loses balance, and both people take a tumble. Each incremental step increases the probability of someone losing balance by some small amount, and it's just random chance whether or not that loss of balance happens. There were a few times during the Cold War when random chance almost got us all killed...
Now, is it credible for someone to say, "gimme $20 or I take a step?" Maybe! Or, perhaps make a bigger demand than $20... Depending on the stakes, the threat to take a step can be credible, even when both people are sane.
That's the concept. Once Schelling introduced this concept to the game theory world, a bunch of math ensued. Two games, in particular, became canonical. War of attrition, and brinksmanship. In brinksmanship, the structure is as follows. Two players. Player 1 decides whether to play or quit. If Player 1 quits, Player 2 gets something. If Player 1 plays, "Nature" decides whether or not to create a disaster, which hurts both players. "Nature" isn't a strategic actor. Nature just rolls dice, and Einstein can cringe all he wants.
Come to think of it, from now on, whenever I insert "Nature" into my games, I'm going to rename it, "Einstein," with a footnote about the legal paperwork it has submitted to change its name to "Heisenberg."
Anyway, if the disaster doesn't happen, Player 2 decides whether to play or quit. Then,
War of attrition? Similar. Players alternately decide whether to play or quit, with each decision to not quit imposing a direct cost.
Both games are repeated games, as opposed to a one-shot, simultaneous move game, like the prisoner's dilemma. The way we solve these games is with a technique called "backwards induction." The math is sometimes ugly, but the concept is simple. You start your logical process at the end of the game, and work your way back to the beginning. Consider brinksmanship. One player, for whatever reason, may have more tolerance for risk than the other. Why? We can build that into utility functions. It's just the nature of preferences. There are plenty of reasons, but for whatever reason, one player will be willing to step farther down that hill than the other. (Further? It's a metaphorical hill, so... But I am speaking of a hill. Farther. I'm going with farther here. Distance, movement... farther. I don't care that the hill is imaginary and metaphorical, I'm going with "farther.") What does that mean? Let's say Player 1 will step farther down the hill than Player 2. If it gets to the round where Player 2 would quit, then Player 2 has just run the risk of a disaster. Before that, Player 1 will definitely have played. If there wasn't a disaster before that. And so on, back through to the beginning of the "game tree." If Player 2 knows that Player 1 will definitely go farther down the hill, then the rational move for Player 2 is to quit as soon as it is Player 2's turn to make a decision. Otherwise, all Player 2 is doing is running risk after risk of disaster, and either facing actual disaster or quitting after facing pointless risk. War of attrition gets solved the same basic way. The one who will quit in the end should quit as soon as it is possible.
Backwards induction. This is at the core of modern game theory, and goes by the fancy-sounding term, "subgame perfect Nash equilibrium," when you work out all the math with a bunch of fancy equations 'n stuff. Sometimes, you need probabilities, 'n stuff, and our boy, Einstein will get uncomfortable, but hey. Try running an actual experiment sometime, Al! Oops! Did I say that?
I still love you, Al!
Anyway, what does this mean, for strategic interaction? Simply, this. How does it all end? Work your way backwards.
Impeachment. This is an absolute, sure-fire loser. The reason Tom McClintock double-dog dared Democrats to impeach Trump is as follows. Suppose Nadler brings articles of impeachment. Here are the possible sequences. It goes to the floor of the House, and they fail to get a majority of the House. Why? Not all Democrats are on board! Why not? Mainly, a lot of them think (correctly) that it would backfire in 2020. So, Nadler loses on the floor of the House. Or, the articles pass on the floor, and Trump is "impeached." Next? Trial in the Senate. Absolute, 100% chance of acquittal in the Senate. I'm not even going to play around with the statistician's bet-hedging here.
Trace that back. It means that the Democrats either have to pull back from the impeachment "proceedings," or lose. Nancy Pelosi gets this, and has been telling the party not to pursue impeachment because she's doing "backwards induction." A lot of the party, though, isn't. They're just looking at the next move. That's how you make the wrong move in game theory.
Let's turn, now, to Corey Lewandowski, and in fact, the whole parade of witnesses brought in to "testify" before congressional committees. Trump has told his people not to answer any questions. Can he do that? In practical terms, yes. Legally? Um... the boundaries of executive privilege are...
Anyway, what happened with Lewandowski was 100% predictable. He was instructed by Trump to stonewall, so that's what he did. Was he in contempt of Congress? A pretty strong argument can be made, and everything about what he did was predictable. In advance.
So, what could Nadler have done? Inherent contempt. Simply voting to hold Lewandowski in contempt wouldn't do anything. Wag your finger... whatever. If the goal is to ask the DoJ to intervene on Congress's behalf... good luck with that. However, there is this little thing called, "inherent contempt." Congress declares a non-cooperative witness to be in "inherent contempt" of Congress, has its Sergeant at Arms arrest the witness, and throw that witness in Congress's own, little cell. Yes, it is sort-of possible for Congress to have someone arrested rather than having the executive branch do it, when they declare inherent contempt.
But... does that sound kind of risky to you? Like, maybe we don't know how that would play out?
A little... brinksmanship-like?
But here's what I did know from the beginning. Nadler would never run the risk of trying. He would never step that far down the hill. Lewandowski, on the other hand, absolutely would stonewall. That made the outcome of Lewandowski humiliating the Democrats by stonewalling and getting away with it absolutely inevitable. Just as every other Trump official and former associate has succeeded in doing. Remember when Matt Whitaker started calling "time" on the Democrats questioning him? In every one of these hearings, we know what will happen. Trump's people will refuse to answer any questions. They will stonewall, and run down the clock. And, since the Democrats won't use inherent contempt due to their unwillingness to run the risk, that means the act of bringing in a witness like Lewandowski is pursuing a losing strategy.
Democrats aren't applying the logic of backwards induction. If they were, they wouldn't bother with any of these hearings. That doesn't mean hearings can't be useful, but if you know a witness won't cooperate, and that the DoJ and courts won't cooperate, then taking inherent contempt off the table means that hearings only matter if you have cooperative witnesses.
So, what do Republicans do? Here's the thing. When someone doesn't pursue a course of action because they've done the backwards induction, then by definition, you rarely notice. It's a dog that doesn't bark, much of the time. It isn't that the GOP just kind of rams through everything. Instead, consider the following.
2017-8. A Republican President, House and Senate. What didn't happen that Republicans would have wanted? What didn't they even try? How about privatization of Social Security? Just as an example. This has been floating around in the undercurrent of the conservative movement for decades, but every time it rises to the surface, the GOP gets slapped down hard. By backwards induction, what would have happened if they had tried in 2017 or 2018? Well, it would have failed because they couldn't have gotten a Senate majority, and then the unpopularity of the proposal would have hurt the party in 2018. So, they didn't do it. They didn't even try. That's backwards induction.
And because the party used backwards induction, unlike the Democrats, you didn't even notice. The dog didn't bark. Now, you may be thinking that the party still lost the House, and there are plenty of factors, but what would have happened if they had pushed on Social Security privatization? It would have been far worse for them. The value of practicing backwards induction is not facing that consequence.
On the other hand, the Democrats keep walking into losing situations by failing to trace out the obvious implications of their decisions. These hearings... we know how they go. Uncooperative witnesses will stonewall, and the hearings are pointless. The Democrats will not be able to remove Trump through impeachment. Trace this out. The math isn't that hard.
And read Thomas Schelling.
And for those still finding their way to the new blog, welcome, and spread it around.
Comments
Post a Comment