When they change their minds it’s called “flip flopping”, when we do it it’s called “updating on evidence”.
As a rationalist, it bugs me to no end that changing one’s mind is considered a sin for politicians when it should count as a great virtue for everyone. I don’t know if it helped or hurt Hillary Clinton’s election chances, but it was painful to watch her pretend to have always supported gay marriage when she and her husband clearly opposed it in the early nineties. Would anyone have held it against her if she said that she spent a decade listening to LGBTQ folk describe their struggles and they changed her mind? I have seen people in my life update their attitude on gay rights and admired them for it, I would be proud to support a politician who did that.
Since I’m utterly unelectable as it stands I can admit freely: I change my mind all the time. Almost certainly, I don’t do it often enough. To encourage this process, the least I can do is get a blog post out of it whenever I flip flop. Here’s part 1, with many more hopefully to come.
Flip Flop #1 – I no longer believe that the FDA harms more lives than it saves
So put your matches away until further investigation. A lot of commenters on “EpiPenomenon” and on Reddit disparaged me for being a “101 economist” naively applying simplistic models to complex subjects. These comments did absolutely nothing to convince me. They neither offered a better model, nor told me something I did not know about drugs, nor refuted Economics 101 – that incentives drive decisions. Ultimately, these are genetic arguments – attacking the presumed source of my belief (a blind faith in markets) instead of the arguments themselves (that the FDA is incentivized to err strongly on the side of approving too few drugs too late).
What changed my mind was my mom, a chemist working in the pharmaceutical industry, who explained in detail the rules the FDA plays by and how they go about their job. In particular:
- The FDA has fast track programs with hard approval deadlines for important breakthrough drugs.
- The FDA relies on a global network of independent testing laboratories, which are subject to audit. These audits include FDA people physically looking over people’s shoulders and nitpicking their procedures. This network of labs requires a consolidated regulator enforcing a single standard, it wouldn’t work as well with the multiple competing xDAs I proposed.
- I underestimated the damage historically caused by untested drugs, the worst of which do their damage over a long period of time (e.g. through birth defects). This prevents a quick public reaction to harmful drugs that would limit the number of victims.
There was clearly a lot of direct evidence regarding the FDA I was unaware of, but many people found it easier to assume that I’m a brainless libertarian than to google any of it. The hierarchy of ways to change my mind on an issue is quite clear. Please keep the chart below handy when trying to convince me of something:
- Direct evidence
- Large bribes
- Circumstantial evidence
- Modest bribes
- Genetic arguments
Flip Flop #2 – I no longer believe in a weakish form of the Efficient Market Hypothesis, namely that prolonged market distortions are limited by the amount of money at stake
Election night didn’t change my mind about what is possible in politics. Aside from having a different president, a country of 49% Trump voters is very much the same country as one with 51%. But a core belief of mine was shaken on election night: the efficient market hypothesis.
In the long-gone days of my carefree youth (2009-2010), I worked as a day trader for a hedge fund. My job consisted of reading economics news and clicking “buy” and “sell” as asset prices danced a merry jig on my five PC monitors. Then I read Nassim Taleb’s Fooled By Randomness, learned about efficient markets, realized that my trading returns are indistinguishable from random noise, quit my job and came to the US. Needless to say, the efficient market hypothesis played quite a role in my life.
A lot of my smartest friends work for hedge funds as quants and traders, and I understood that markets may be mistaken about a 0.1% discrepancy in the price of pork belly futures in Frankfurt and Chicago. Catching these mistakes is worth enough to keep my friends fed and clothed, but the amount of these mistakes is limited by the profitability of catching them.
I still believed that the markets are generally efficient with regards to common knowledge – that the current price of any traded asset (like a stock) completely reflects everything that is known to the public regarding that asset. I fully believed that all the way through business school, where every single finance professor endorsed weak form EMH and where I rejected several opportunities to go back to the investment management industry. I kept absolutely believing in weak form EMH up until 4 am on election night:
At least since October, the US stock market projected that a Trump victory would drop American stocks by 10%-12%. This wasn’t hot air spewed by talking heads on TV, it was reflected in the actual behavior of stock prices, in every buy and sell by a trading desk in Hong Kong or an algorithm in Greenwich. For this projection to be wrong meant that there was a pile of 2 trillion dollars, twice the GDP of Russia, free for the taking. The efficient market hypothesis doesn’t rely on all market participants being sane or even a majority of them, just enough sane people to grab 2 trillion dollars of free money.
The 10%-12% projection more or less matched the stock market behavior on election day, with stocks rising 2% as Trump’s chances dropped from 30% to 10% in the early afternoon, then crashing in the evening as Trump’s victory became likelier and likelier. And then, at 4 am, all the humans and robots who until that point believed that “Trump = 10% drop” changed their minds simultaneously and US stocks hit historical highs.
I have read several explanations of this, and they are all obviously fake in the sense that believing the “explanation” and knowing the election results ahead of time would not have caused you to predict the stock market movement that occurred. The stock market could not rise 10% on news of a Republican senate because the senate races were decided by 11 pm and the odds of a Republican congress given a Trump win were around 90% anyway. The stock market could not have risen on confirmation of a peaceful transition of power because Hillary was winning the popular vote and there’s no way that massive protests seemed less likely at 4 am than the day before.
A lot of very smart people built very complex models that had to account for all eventualities, from a recount in 5 states to a terror attack the morning of the election. The actual outcome (Trump win, Republican senate, Clinton concession, conciliatory victory speech) landed well within the anticipated range. And then, suddenly, everyone decided that the models were wrong after all and Trump is good for US stocks.
This leaves us with three options:
- Everyone trading US Stocks was collectively insane in their projection of the election’s impact, and there were indeed 2 trillion dollars up for grabs for almost a month, up until election night.
- All the models before the election were correct, and everyone is insane right now due to runaway optimism bias. This means that there are 2 trillion dollars available this very moment to anyone shorting US stocks.
- Everyone was insane, is insane, and in fact the stock market is not driven by the correct pricing of publicly available knowledge but by restless voodoo spirits flitting to and fro likes yo-yos in the hands of capricious gods.
Take your pick, and let me know if your hedge fund is hiring.
22 thoughts on “Flip Flops, part 1 of ∞”
First hypothesis is correct. And not so strange that people would be collectively insane in that way. People are typically insane where politics is concerned. So saying “stock market is going to drop” was just a way of saying, “Trump is really bad.”
And not so strange that they figured it out at 4:00 AM after the election. A few days before the election, someone said there was a 99% chance that Hillary was going to win. I said, fine, I’ll give you $10 if Hillary wins, and you give me $1,000 if Trump wins. They said no, that would be break even, so no reason to accept. I said, fine, $10 vs $500. They still wouldn’t accept. And why not? Because the claimed odds was just political posturing.
Same thing here. If people continued to insist that a Trump presidency would have awful financial effects, after it was a reality, they would start losing money by being wrong. So they stopped being wrong, in order to avoid losing money, just like the fellow who wouldn’t make that bet with me (really sad he didn’t!)
Your story contradicts the hypothesis: your friend was aware enough before the election that he didn’t want to bet on Hillary. No one does “political posturing” by programming their models to sell stocks when a district is counted for Trump on election day. Those that did program their models this way did in fact lose billions and billions of dollars.
Ok, so here’s an adjusted model. There are two groups of people.
Group 1. These are the people who originate the story that the stock market will drop. These people are engaging in political posturing. They think that Trump is really bad, and this explains why they tell this story, but they do NOT believe the stock market will drop, and do not plan to bet on that, just like the person who wouldn’t bet on Hillary.
Group 2. Ordinary people. These people believe Group 1. So they actually believe the stock market will drop. Since these people are much greater in number than group 1, the stock market actually acts like this is going to happen.
Then on election night, the pushback from group 1, who are now making money, causes group 2 to come to their senses rather than losing even more.
Does that work?
I still don’t think that works, because most of the money in the stock market is invested by Group 3: funds and institutions. Banks, pension mutual funds, hedge funds… this is the vast majority of the money and it’s controlled by finance professionals and quant models. These people are smarter than the NY Times (Group 1) and have more than enough money to immediately correct any price movement by individual investors who read the NY Times (Group 2). This doesn’t mean that 90% of Group 3 can’t fall under the same mass hypnosis, but it’s still jarring to realize that something like that probably happened.
LikeLiked by 1 person
Don’t those models often swing into a bubble for a bit before they pop it?
I feel like there are still enough “rich idiot” investors to start to create a stock market problem of some sort, which might take a full six hours to correct.
“How efficient is the efficient market” is a very interesting question that I’m not sure anyone is investigating. It’s clearly not perfectly efficient, and it’s clearly not perfectly inefficient, but nobody seems to care how efficient it is despite this being the central point of contention between eg social democrats and (pragmatic) libertarians.
Every hedge fund in the world in investigating that precise question.
Anyway, the “Trump = -10%” held for a month, not just election. You could see clear correlated and simultaneous movements of Trump’s odds and the stock market on events like the FBI investigation announcement. I agree that the anti-Trump price movement had many characteristics of a bubble, but bubbles usually pop with the accumulation of new information, not suddenly in the middle of the night. For example, the housing bubble started popping when data on mortgage default rates and construction started looking horrible. In retrospect the information was there ahead of time (and some people saw it), but the evidence that houses/mortgages are losing value kept accumulating until the delusion finally broke.
I guess you could be right about a bubble: there was a lot of evidence that Trump would be a generic and not completely incompetent Republican president (as opposed to “literally Hitler”) and that is not bad news for US stocks. The victory speech was the final piece of evidence that broke the spell.
Well… the appearance for not a small number of people is that Secretary Clinton only started supporting gay marriage when it became politically advantageous to do so. This suggests that her consideration on the issue is based on what benefits her personally as opposed to principle (and people claim to want principled politicians). The insistence on being on the side of queers all along as opposed claiming to having undergone a conversion fits in with this view. This flip-flopping garners less respect then those who describe some credible method by which they changed their opinion.
(to Hillary Clinton’s credit, it was her husband who signed the Defense of Marriage Act and wifes don’t believe everything their husbands believe. Sure Mrs. Clinton could have spoken out against it at the time but there may have been other priorities like denouncing supper-predators. It is conceivable that she has been in favor of queer rights this whole time just didn’t say much about it until the political costs dropped bellow certain point. This is better though by how much is… questionable. Also, both Clinton’s have reputations for being unprincipled and it’s easier to fit in Secretary Clinton’s actions into one’s established narrative then to give her the benefit of the doubt.)
In general, accusations of flip-floping in politics are supposed evidence of lack of principles. There are politicians who try to explain changes in mind as principles to varying degrees of success (Senator Orin Hatch on AIDS aid to Africa for example). Your changes in opinion fits the narrative I have for you so it is (mostly) unsuspicious and a reaffirmation of your stated principles. Unless you’re receiving bribes from the FDA to make your change of heart sound legitimate, I believe that you’ve changed your mind for principled reasons.
Alternative Theory: Some shadowy figure involved in running things behind the scenes said something to reassure world banks – for example, a senior figure on the trump campaign called the manager of a big bank and said “listen, we’re serious about deregulation and infrastructure, but the parts about being anti-trade-deals are just posturing.” (Or more generally, some new piece of information went out to banks that the general public is unaware of).
I’d generally consider this sort of conspiracy unlikely, but as you said, the alternatives are also pretty weird.
I would actually love to live in a world in which competent figures are running things behind the scenes, but all evidence points to their nonexistence. Unfortunately, it’s Trump all the way down :-/
The WEAK form of the EMH doesn’t say that no distortions (meaning deviation from fundamentals) is possible. That is the strong EMH. The weak form just tells us that the current price (module discounting etc..) has to be the expectation of the future prices (ok not quite…really definied in terms of martingales but fine).
What you are missing in the election is the chance that something went WAY worse. Trump has been mostly sane, using relatively experienced operatives and backed away from many campaign promises. There was always a small chance that he got elected yelled “WOO HOO” immediately announced his intention to scuttle/withdraw all our major trade deals (congressional approval or not…I believe he can withdraw on his own), Indeed, my understanding is that his team has been meeting with buisness interests in a reassuring way fairly consistently. He could have announced that he intended to appoint a special prosecutor to go after Hillary and we could have ended up on the brink of a constitutional crisis. He might have announced that he would withdraw us from NATO.
None of these things were likely, all very small probabilities, but that would have had massive impacts on the stock valuations and we observed they didn’t happen.
I’ve heard those explanations, but the math still doesn’t work out. How bad would withdrawing from all the trade deals AND prosecuting Hillary would have been? Would it be enough to wipe out one quarter of all future profits of US companies forever? Unlikely, but let’s grant that these disruptions would cost 25%. And what was the likelihood of that happening, before and after Trump’s speech? It couldn’t have been likelier than 20%, and let’s say a single speech halved these odds to 10%. A 10% reduction in the odds of a 25% drop is 2.5%, that’s one fifth of the difference between how the market priced Trump at 9 pm and 5 am. For this explanation to account for more than a small part of the shift, you need quite implausible numbers on the odds of Trump setting the economy on fire, how much these odds could drop in a few hours and how much long term damage he would’ve done.
No, because you can’t assume all of the risk doesn’t merely comes from what Trump might do but also how people in the financial markets would react to Trump’s behavior.
I suspect a substantial portion of that loss comes from the effect on future prices that would result merely from a larger loss of confidence by other investors than predicted. The most important (and often hard to estimate) factor of future value is the future attitudes/optimism of other investors.
What the markets learned wasn’t merely the content of Trump’s speech but the reaction of other market participants to the election and the speech and how Trump’s election affected their inclinations to make certain kinds of inferences. Given that the market value would drop instantly to 0 the moment that we believed it would drop to 0 in 10 years or 100 years (the expectations push back…at 9 years everyone will pull out and so forth) even small shifts in pessimism/optimism (not great descriptions) can have outsize effects on the market valuation.
While you claim to be talking about the weak EMH you seem to keep acting as if you are talking about the strong EMH. The weak EMH doesn’t require that market valuations be particularly well tied to states of affairs out in the world. It can accommodate behavior where market value fluctuates pretty wildly as a result of slight biases in the procedures we use to estimate others likely future behavior. I keep meaning to work out the math and see if/when it is compatible with actual bubbles as well. Just because the price tends towards infinity doesn’t mean the expectation of that price at t + delta t isn’t the price at t for any finite delta. Point is that weak EMH just doesn’t require the kind of common sense market behavior you seem to assume it does.
Besides, I don’t see how any argument against the weak EMH that runs as “but people could have bought” can be convincing given that they didn’t in fact buy in massive quantities. As long as you believe there are some decently smart people in finance their failure to buy isn’t evidence against weak EMH but against the likely truth of your analysis.
I was talking about the simpe versions of EMH: weak = can’t beat the market using historic prices, semi-weak = can’t beat using public knowledge, strong = can’t beat even with private knowledge. I assumed the semi-weak version, and the implication that current market prices reflect the best prediction given all public knowledge. If you were thinking of more technical definitions, I’m sorry I’m not on your level :)
Anyway, I think you’re right in that the attitude/optimism of other market participants isn’t shared knowledge, and perception of it can fluctuate quickly. I know it’s silly to generalize from personal evidence, but I couldn’t help but notice that the market reflected the emotional state of me and a lot of my friends who are generally libertarianish but rooted for Hillary this time around: 11/8 morning – can’t visualize a Trump win, 11/8 evening – depression, 11/9 – optimism kicks in again. If a lot of investors went through the same emotional states and immediately noticed each other doing the same thing, that would account for the market behavior.
To me, this progression represents a failure of rationality and imagination. I guess I expected “the market” to be smarter, but it’s also just made up of nothing but silly people :)
Well, first off the weak form almost certainly does not hold — at least not on short time scales where algo trading based purely on microstructure (“price”) information dominates, My intuition is it holds even worse on longer horizons but that is hard to prove.
In this specific instance, I would chalk it up to different players involved at different times. Leading up to the election, responding to polls and probability shifts, I would expect mostly algo-driven flow to be in control. Once the result came in, real money (major end users with longer horizons, not short term traders w/ models) is what moved the market. Basically, predictive models were underestimating the amount of real money buying that was to be triggered on Trump win. e.g. http://www.bloomberg.com/news/articles/2016-11-09/icahn-left-trump-victory-party-to-bet-1-billion-on-u-s-stocks
You keep saying ‘2 trillion up for grabs,’ are you exaggerating for affect? There certainly was not 2 trillion up for grabs. That would imply 2 trillion was ‘grabbed’ by someone else and was transferred, when it certainly wasn’t.
I’ll wait for a JF paper to really map this stuff out. But I’ve done some academic research on this stuff before, and the big issue here is liquidity. The major drop was in a less liquid after-hours futures market.
As an example of what could have happened: Trump wins. A set of investors panics. On after-hours future markets there isn’t much liquidity, and aren’t a lot of market makers actively trading (I am no expert on after-hours market dynamics, hence why I’ll wait for an academic paper to clear that up). So let’s say Hedge fund A wants to offload $50mm of SP500 futures. Well, since it’s not a particularly liquid market they are able to offload the first $10mm at a relatively good deal. But then the next $30mm is at a 3% discount. By this point they still have their sell order in, but liquidity is dry. In addition, a market maker algo is detecting a rapid drop in the price which sends off a warning signal. The question: Is this signal reflecting a state of the world they are not familiar with? or is it reflecting a state of the world they are familiar with, but have decided they aren’t worried about? Unclear. These things take time to compute and resolve for thoughtful humans. Sometimes they take a few hours to compute.
The final $10mm is waiting on the open market, and finally they sell it at a 5% discount. This sale is now the latest bid/ask measure. The implication is that $2 trillion dollars of market cap was wiped out. However, the true market cap is a mapping function of marginal investor market beliefs to the market, but this can be out of sync temporarily. Most likely it will be out of sync when a huge amount of new information is needing to be digested, computed, OR there are structural issues (liquidity, after hours) messing up the information set. If you wanted to buy the market bigly during that dip, you probably wouldn’t have been able to. Liquidity issues and after-hours is not a friendly environment.
Another way to think about it: The weeks leading up to the election reasonable expectations of Trump winning were 10-40% (take your pick, doesn’t really matter). Now, if Trump were a risk to the market. A 2 trillion dollar risk. That should have been priced in already. It wasn’t. We see the market was indifferent, or at least not different enough for smart econometricians to agree on a clear measured signal. This means all the HFs around the time were saying “Yeah Trump very well could win, I’m not particularly interested in changing my portfolio as a result.” The only clear signal was the US/Peso ratio.
So, plausibly, a few funds or people freaked out, sold, it was after-hours, there were liquidity issues, the price goes back to normal when the market is again representing the marginal investor in a liquid market.
I’m not telling you that is what happened, but it could be the case (and probably is the case). So it’s sort of silly to say there were “$2 trillion up for grabs.”
Thanks a lot for the informative comment. Liquidity limitations and market reaction to big buys/sells are definitely part of the story, but I don’t think this dispels the main thrust of my argument.
First of all, the $2 trillion is the total market cap gap, so it’s obviously an exaggerated figure for any practical purposes. No one could have actually made a trillion dollars trading the election. However, the money to be made was not just on election night, it was in the months before.
The market didn’t drop just after the Trump win, the projected drop reflected his chances rising and falling for weeks. A predicted drop of 10% was a consensus I have seen in a lot of places, and it was priced in for example in the market reaction to things like the Clinton FBI letter on 10/28. Bloomberg has a good chart and analysis of that. There was plenty of time after Trump’s chances rose (and the market fell) on the FBI letter to buy low. Perhaps not quite $2 trillion, but easily hunderds of millions could easily have been made (daily trading volume on just the SP500 index is in the ~billion order of magnitude). Some people probably did make a lot of money by buying low when Clinton dropped, but there weren’t enough of them to correct the market before November 9th.
According to your investopedia link, the weak form EMH “claims that past price movements and volume data do not affect stock prices…future securities’ prices are random and not influenced by past events. Advocates of weak form efficiency believe all current information is reflected in stock prices and past information has no relationship with current market prices.”
Wikipedia’s article on EMH has a weak form section that states: “In weak-form efficiency, future prices cannot be predicted by analyzing prices from the past. Excess returns cannot be earned in the long run by using investment strategies based on historical share prices or other historical data. Technical analysis techniques will not be able to consistently produce excess returns, though some forms of fundamental analysis may still provide excess returns. Share prices exhibit no serial dependencies, meaning that there are no “patterns” to asset prices. This implies that future price movements are determined entirely by information not contained in the price series.”
I quote these definitions because I am confused. How is weak form EMH violated by the stock market’s movements and the election results that you discussed? Seems like weak form EMH is violated if you can analyze past stock prices to consistently produce excess returns (by predicting future prices better than chance). I don’t see where you argue that an analysis of past stock prices could have been used to make money in this scenario.
You seem to be arguing against one of the stronger forms of EMH, as far as I can tell.
Why must the same group of actors be responsible for both market movements? Imagine an alternate scenario where some company announces some new product. First, their stock price goes up as some investors feel the company’s financial outlook has improved and buy some shares. Second, the stock price goes down a bit (but is still above the pre-announcement price) as some pre-existing investors sell some of the company’s shares to rebalance their portfolios. No one in this tale changed their mind. We just had different people doing different things at different times.
I was talking about public information and not just historical prices. So you’re right, I should have said “semi-strong EMH”. With that said, SSEMH is also widely believed by most finance academics and by me. Only strong form EMH (that even insider trading can’t get excess returns) isn’t consensus.
If we do accept SSEMH, the price of the company shouldn’t go down when the original investors want to rebalance. Profit seeking investors should be happy to buy it at the high post-new-product price, or just below it, since the underlying information about the company’s profitability doesn’t change.
In any case, what are the chances that a bunch of people decided to rebalance their portfolios by selling the entire US stock market in perfect coincidence with the CIA Clinton letters coming out?
Thanks for the clarification. Also, now that I’ve read the other comments more carefully, I see that some one else made the same question. Hopefully an edit to your original post will prevent additional people asking about which EMH form you meant.
Exactly how is semi-strong EMH violated in your Trump scenario? Are you saying a person could do fundamentals analysis during similar elections and get excess returns more often than not? As far as I can tell, the market being efficient (of the semi-strong EMH variety) doesn’t mean the market makes perfect predictions, just that humans can’t reliably predict better. But perhaps I am misinterpreting the definition of semi-strong EMH.
I agree that the US stock market seems to be efficient enough that it is very hard to try to make money off of predicting these rebalances. It was just a tale about how a price can be tugged in different directions by different people acting for self-consistent reasons with no one changing their mind.
My hypothetical scenario was about a single company’s stock price after announcing a new product. I wouldn’t get hung up on the particular details I presented there. There are many other possibilities.
(Do you mean FBI letters rather than CIA letters? Also, there are two letters of note, right? One that decreased Clinton’s chances and one that increased them, right? I might be misinformed as I do not follow the news closely. Thanks for any corrections.)
Stock market movements correlate with a lot of things, and the market always seems to be moving. A two-time correlation criteria would be met by a huge number of theories about the stock market, many of them incorrect/inaccurate. Also, are you attributing all of October 28’s and November 7’s movement to the letters (I say this based on the annotated plot at one of your linked articles)? That seems a tough argument to make, as again, the stock market correlates with a lot of things and does plenty of movement even when there aren’t letters being made public. Also, it seems a lot of work would have to be done to nail down what the market “thinks” Trump’s chances were at each time, as the market might not “agree” with your chosen polls.
So, how likely is it that we would get a coincidence of the stock market movements with letters being made public? My very rough estimate is somewhere between 0.25 and 0.10.
Are there other observations that lead you to your conclusion about what the market “believed” about a Trump presidency? If so, I think that could help bolster your claims a great deal.
Tip: update the old FDA post to link to this one.