On uncertainty: Recessions, impeachment, trade... take your pick
There are a lot of ways I could introduce the general topic of the morning, which is another step back... step back...
Will there be a recession by the end of 2020? The betting market over at PredictIt currently puts the chances of that at around 40%. Close to 50-50. An impeachment? The gamblers... excuse me... investors call that 70% as of this morning, with "yes" shares trading at 70 cents on the dollar, but I'm a little more circumspect. Over 50%, but I'll claim more... "uncertainty." Trade? We just got something that Trump is calling "phase 1" of a trade deal with China, but mostly it is ratcheting down of some tariffs, and calling off some new ones. We don't really have settlement on, oh, things like intellectual property issues, which nobody with a brain really thought would be settled. What's really going to happen, long term? No clue.
Now, when I say, "no clue," what does that mean?
I think, maybe, it's time for some musing on probability theory. I frequently write about "Bayesian" interpretations of probabilities. If I say there is a "40% chance" of a recession by the end of next year, that means I don't have full information about the economy and everything that has or will happen. Given the state of the information I have, .4 is a sort of relative assessment of recession indicators and contra-indicators, and their relative strengths. I aggregate the information I have, and .4 is that aggregation. Either a recession will happen, or it won't. I just don't have full information, so I summarize my information with something that I call "a probability." And, as I get more information, I update that probability.
In that interpretation, a probability of .5, or, 50-50 means that the information suggesting that an event will happen and the information suggesting that an event won't happen are equally compelling. So, I am maximally uncertain. As far as the likelihood of a recession by the end of next year goes, that's close to where we are. Why? The bond yield keeps inverting, trade tensions, and a few other things. So, I "don't know" if a recession will happen by the end of next year, and of course, that has political implications. That's one of the reasons Trump keeps haranguing Powell to drop interest rates to zero. He's terrified of a recession.
I have somewhat more information about congressional Democrats, so I have somewhat more certainty about the likelihood of impeachment, but I still lack a lot of information, both about what will be revealed in the future, and how the Democrats are deliberating internally, hence I summarize my information with a number that has some proximity to the 50-50 line, and I will continue to "update" my assessments as new information becomes available.
There is another interpretation of probabilities, though, which is that they reflect actual, random processes that can only be characterized by probabilities in a universe governed by non-deterministic processes. If there is anything in the universe that isn't governed by deterministic laws, it'd be subatomic particles, or... us. Every time some lousy driver does something unpredictably stupid on the road, just think to yourself: only a subatomic particle could be that unpredictable! Or, obviously, some teenager, texting-and-driving. So, maybe humans just behave with some element of randomness. And we characterize that with probabilities. The mathematics of randomness.
Hence the ideological debate raging within statistics departments across the world.
However, within either of those conceptions of "uncertainty" is some aggregation of information. However we conceive of a probability, part of the reason we are willing to put a probability on a recession by the end of 2020, or an impeachment is that there is information. The process of putting a probability on a recession is, perhaps, more quantitatively grounded given that we have a longer history of economic data, more recessions, and the capacity to run statistical models that estimate the likelihood of a recession given any set of circumstances, whereas there have only been two past impeachments in US history and there is no precedent for anything like the current situation, but in either case, we are working with direct evidence and knowledge.
When I said that I have "no clue" what will happen with China, is there a fundamental difference between that and saying that the odds of a recession are close to 50-50? I think so. At one level, sure, I didn't give a specific event to say either will or won't happen, but OK, let's resolve that.
Will Trump and China reach a "phase 2" deal with real policy resolutions by the end of 2020?
My answer? Still... "no clue."
The short version of why: A few months ago, I would have said, not bloody likely. What changed? Trump is now under more pressure because of Ukraine, and tension within his own party on Syria. To quote Londo Mollari from Babylon 5, "Only an idiot fights a war on two fronts. Only the heir to the throne of the kingdom of idiots would fight a war on twelve fronts."
Without comment on any modern-day analogs to Cartagia, the point is that as the number of fronts goes up, the pressure to stop fighting on at least one of them also goes up. That just happened. The "phase 1" announcement was Trump trying to pull away from the China trade war because he is too embroiled in too many other fights. How does that play out, eventually? Since it is so contingent on stuff I don't know, it is something that I don't even know how to estimate.
And there's a fundamental difference between saying that, "A and B are equally likely," and, "I don't even know how to estimate the likelihoods of A or B."
Colloquially, we'll describe each state as, "uncertain," but in terms of statistical theory, and how scholars approach the question, they're very different.
Will there be a recession? Will Trump be impeached? I can put numbers on answers to those kinds of questions because I have information. The information just runs in opposing directions, and sometimes isn't very strong.
Then, there's the stuff on which I just have no information. The world is increasingly filled with those kinds of questions.
Uncertainty can mean a lot of different things. There's a lot we don't know, and there's a lot that we can't even estimate. That doesn't mean 50-50.
Will there be a recession by the end of 2020? The betting market over at PredictIt currently puts the chances of that at around 40%. Close to 50-50. An impeachment? The gamblers... excuse me... investors call that 70% as of this morning, with "yes" shares trading at 70 cents on the dollar, but I'm a little more circumspect. Over 50%, but I'll claim more... "uncertainty." Trade? We just got something that Trump is calling "phase 1" of a trade deal with China, but mostly it is ratcheting down of some tariffs, and calling off some new ones. We don't really have settlement on, oh, things like intellectual property issues, which nobody with a brain really thought would be settled. What's really going to happen, long term? No clue.
Now, when I say, "no clue," what does that mean?
I think, maybe, it's time for some musing on probability theory. I frequently write about "Bayesian" interpretations of probabilities. If I say there is a "40% chance" of a recession by the end of next year, that means I don't have full information about the economy and everything that has or will happen. Given the state of the information I have, .4 is a sort of relative assessment of recession indicators and contra-indicators, and their relative strengths. I aggregate the information I have, and .4 is that aggregation. Either a recession will happen, or it won't. I just don't have full information, so I summarize my information with something that I call "a probability." And, as I get more information, I update that probability.
In that interpretation, a probability of .5, or, 50-50 means that the information suggesting that an event will happen and the information suggesting that an event won't happen are equally compelling. So, I am maximally uncertain. As far as the likelihood of a recession by the end of next year goes, that's close to where we are. Why? The bond yield keeps inverting, trade tensions, and a few other things. So, I "don't know" if a recession will happen by the end of next year, and of course, that has political implications. That's one of the reasons Trump keeps haranguing Powell to drop interest rates to zero. He's terrified of a recession.
I have somewhat more information about congressional Democrats, so I have somewhat more certainty about the likelihood of impeachment, but I still lack a lot of information, both about what will be revealed in the future, and how the Democrats are deliberating internally, hence I summarize my information with a number that has some proximity to the 50-50 line, and I will continue to "update" my assessments as new information becomes available.
There is another interpretation of probabilities, though, which is that they reflect actual, random processes that can only be characterized by probabilities in a universe governed by non-deterministic processes. If there is anything in the universe that isn't governed by deterministic laws, it'd be subatomic particles, or... us. Every time some lousy driver does something unpredictably stupid on the road, just think to yourself: only a subatomic particle could be that unpredictable! Or, obviously, some teenager, texting-and-driving. So, maybe humans just behave with some element of randomness. And we characterize that with probabilities. The mathematics of randomness.
Hence the ideological debate raging within statistics departments across the world.
However, within either of those conceptions of "uncertainty" is some aggregation of information. However we conceive of a probability, part of the reason we are willing to put a probability on a recession by the end of 2020, or an impeachment is that there is information. The process of putting a probability on a recession is, perhaps, more quantitatively grounded given that we have a longer history of economic data, more recessions, and the capacity to run statistical models that estimate the likelihood of a recession given any set of circumstances, whereas there have only been two past impeachments in US history and there is no precedent for anything like the current situation, but in either case, we are working with direct evidence and knowledge.
When I said that I have "no clue" what will happen with China, is there a fundamental difference between that and saying that the odds of a recession are close to 50-50? I think so. At one level, sure, I didn't give a specific event to say either will or won't happen, but OK, let's resolve that.
Will Trump and China reach a "phase 2" deal with real policy resolutions by the end of 2020?
My answer? Still... "no clue."
The short version of why: A few months ago, I would have said, not bloody likely. What changed? Trump is now under more pressure because of Ukraine, and tension within his own party on Syria. To quote Londo Mollari from Babylon 5, "Only an idiot fights a war on two fronts. Only the heir to the throne of the kingdom of idiots would fight a war on twelve fronts."
Without comment on any modern-day analogs to Cartagia, the point is that as the number of fronts goes up, the pressure to stop fighting on at least one of them also goes up. That just happened. The "phase 1" announcement was Trump trying to pull away from the China trade war because he is too embroiled in too many other fights. How does that play out, eventually? Since it is so contingent on stuff I don't know, it is something that I don't even know how to estimate.
And there's a fundamental difference between saying that, "A and B are equally likely," and, "I don't even know how to estimate the likelihoods of A or B."
Colloquially, we'll describe each state as, "uncertain," but in terms of statistical theory, and how scholars approach the question, they're very different.
Will there be a recession? Will Trump be impeached? I can put numbers on answers to those kinds of questions because I have information. The information just runs in opposing directions, and sometimes isn't very strong.
Then, there's the stuff on which I just have no information. The world is increasingly filled with those kinds of questions.
Uncertainty can mean a lot of different things. There's a lot we don't know, and there's a lot that we can't even estimate. That doesn't mean 50-50.
Comments
Post a Comment