This is classic Business Insider. They explicitly ignore the context and try to shake you up with a cherry picked quote. Jamie Dimon doesn't think, and doesn't want his reader to think, that October 15th was a once in a 3B year move. Specifically, in the letter, Jamie Dimon 1) never uses the term "flash crash" and 2) spends the previous five paragraphs explaining exactly how and why liquidity dynamics in the fixed income markets have changed.
What he is saying is that the Oct 15th move looks really strange, but not if you actually understand how fixed income market makers are reacting to the Volker rule. Quoting Dimon:
For instance, the total inventory of Treasuries readily available to market-makers today is $1.7 trillion, down from $2.7 trillion at its peak in 2007. Meanwhile, the Treasury market is $12.5 trillion; it was $4.4 trillion in 2007.
What he's saying is that market makers have smaller inventories on absolute terms, and much smaller inventories on relative terms, so when multiple big market participants try to move in the same direction at the same time, there is much less inventory available to soak up demand.
Come on, every finance guy knows to cut off tail risk, yet they still don't. Look at LTCM, with all the brains they had still didn't cut off tail risk and Meriwether still came back and got burnt again. Look at Victor Niederhoffer, got burnt twice the exact same way not cutting off tail risk. Then you get a guy like Taleb profiting off tail risk. Complaining about liquidity crunches does not count as its one of the main parts of the tail risk.
What you're talking about is a hedge. Hedges are expensive. A better strategy to a lot of investors is to "hedge" by trading smaller. Of course the spectacular failures you name were spectacular because they weren't hedged and were done with immense leverage. There is no way looking forward to predict if a hedge that eats 1/2 your returns every year is a better investment than just taking a catastrophic loss every several years. And it's not as simple as taking a 50% haircut on every trade: in reality maintaining a hedge will turn many of your small winners into losers.
I personally guard against the prospect both by staying small (trading with only a small portion of my total account) and by maintaining net short deltas (beta-weighted). Keeping short deltas in a bull market with upward drift also has a cost, but it lets me sleep easier at night.
Normal distribution, arrived at via the central limit theorem, assume that the underlying distributions are independent. They may well approximate independence most of the time, but when extreme events happen, the distributions are correlated, the CLT doesn't apply (not in the simple way it is normally presented), and the overall result is most definitely not normally distributed.
Normal distribution is a terrible model for estimating risk.
The quote isn't the first to confuse bad statistics with some inherent weakness in statistics. Black Swan was a worldwide bestseller based on the same canard.
1) You are assuming they used a normal distribution. Statistics isn't all gaussians.
2) There's probably an appropriate statistical model that wasn't used to come up with the 1 in a billion number, be it some likelihood function, Poisson or Zero-Inflated Poisson, etc... It's up to smart people to understand what statistical model they should probably use.
According to "the Big Short" and other sources, before the 2008 crash, high level executives all over the industry were managing with the metric "value at risk".
The "1 in 3 billion years" quote sounds like exactly the kind of line of thinking that was in the Big Short.
It's up to smart people to understand what statistical model they should probably use.
It's not just that their model is wrong -- and VaR is certainly wrong. It's that NO model is right.
That is, the fundamental problem is that they are trying to quantify the unquantifiable. Black swans aren't quantifiable by definition.
If you think there is an "appropriate statistical model", then definitely read the Black Swan. I'm glad it came up many times in this thread, because that means the point has been driven home. If you don't get it, you need the 300 pages of beratement it gives you :)
> It's not just that their model is wrong -- and VaR is certainly wrong. It's that NO model is right.
See chaos theory or google "irreducible complexity". There certainly is a model that is right - an obvious model would be just take all actors, their holdings, and their actions, that trivial model certainly is right. There are even people who have this model available to them : the stock exchanges. In theory, no-one else has it available, and we should be sure they're not using or selling this model, right (heh) ?
Black swans are quantifiable. For instance, calculating the current debt loads of governments will tell you that interest rates aren't going back up without killing social security or something like a 300% (cumulative) inflation. Since we won't be killing social security, a sudden crash (at least 40% drop in SPY) followed by a short-term (~1-2 years) tripling of prices is coming. Alternatively, there could be lots of sovereign defaults, resulting in 10-20% treasury interest rates for 5-10 years. There. Quantified.
Black swans happen because we, as in humanity, have this need to believe that things won't change. When they inevitably do change, the first thing we do is to use capital to get us back to the old situation. But that hardly ever works. When it fails, it requires exponential use of "capital" and the capital is gone (so no-one is ever getting their investment back). A recent obvious example I saw of this is Abu Dhabi : the idea that you can have a lush, green, modern city in the middle of the desert. This is wrong, it's artificial. And every millimetre the city grows comes at exponentially increased energy expenditure, exponentially increased stress on all the systems that support this city, and on a regular (as in weekly) basis a sandstorm comes and makes it blatantly obvious what happens if those systems would fail for even single week. But trillions of dollars are being invested to change this, because it was almost reasonable to do this for a small oasis with a freshwater source. As expected, this city is burning an ever-rising share of the oil income of the state just to keep existing. Black swan coming up, right there.
I don't think we are using the same sense of the word "model". In this context we're talking about using a Gaussian vs some other statistical model.
What does it mean to say that Black Swans are unquantifiable? Take these questions:
1) What is the probability that the S&P will be below 1000 within a month (bigger than 50% drop)? Is it .1%, .01%, .001%, .0001% ? Nobody knows. It's a sufficiently rare event, with so few precedents, that you can't give a useful or reasonable answer, with say an order of magnitude.
2) What is the value of the next YC batch, in 2020? Well if it has a company like AirBnB or Dropbox, it could be $10 B. Or it could be $100M. Nobody knows. (See Paul Graham's Black Swan farming essay.) Although the entire point of the VC industry is to try to make this return more predictable than it inherently is.
3) On September 10, 2001, if I asked you what is the probability that airliners would crash in to the World Trade Center, what would your response be? There was perhaps a prior in the 1993 World Trade Center attack. Still, nobody (outside the State department) could have spoken meaningfully about the probability of this event.
non normality of asset returns are known to finance in the 60s, maybe even earlier.
the normal distribution is still used in many cases because it makes problems solvable.
in any case I wonder how much of these price aberrations are caused by bad statistics and not exuberance, overconfidence, or system failures. the latter things can't really be captured by math.
It's not just financial events. Various psychometric traits (such as G) are not normally distributed. The tails are much too fat. So you get a bunch of policy decisions based on assumptions that most people roughly average when that's not true.
Wall Street has historically underestimated tail risk. I'm curious under what model this is a "once in 3 billion years" sort of move. Basically, financial markets don't behave as a normal distribution, and black swans are a lot more likely than their 7-sigma probabilities would suggest.
> of course, this should make you question statistics to begin with
This doesn't make me question statistics at all. What it does do is further my belief that those doing financial forecasting have the same issues as meteorologists: incomplete models.
This is not at all surprising. Statistics isn't the right tool for financial analysis... it's just the best tool that we have. Incredibly unlikely statistical events happen all the time in finance.
A great read on an alternative view on the topic of using statistics for finance is The (Mis)Behavior of Markets by Benoit Mandlebrot [1]. It's very well written, basic enough for most to comprehend and first book on finance I read in college (before I went on to major in finance + math).
The aforementioned book has some very interesting notions with Trading Time being my favorite. Basically one can near-perfectly "forge" financial data with fractal objects called "financial cartoons" [2]. The objects are composed to two distinct fractals - one for price vs trading time and another for trading time vs clock time [3]. The latter rescales the volatility seen former, either compressing or expanding it. Rescaling volatility isn't a new idea, but it was a parallel "discovery".
There has been some work on figuring out how to use Fractal Geometry to analyze financial time series data but it's still in its infancy. The problem is figuring out how to transform the data into the fractals domain + figuring out what the results from a fractal based analysis would mean for forecasting future events. I've been working on these problems for many years (in earnest in college and as a hobby thereafter) but made little true progress.
I suppose that means the Swissie unpegging from the Euro qualifies as a once-in-six-trillion-year event ;)
Could any of the experienced traders on here offer some insight into taking advantage of the upcoming volatility in intra-day Treasuries price-action? Is it mostly done via TY futures and options or are some of the directional, leveraged ETFs (e.g. $TYO) also attractive? Or are indirect strategies preferable?
Unless your timeframe is strictly intraday, I'd avoid the ETFs. In most cases the futures are trading in backwardation which just means that next months futures contract is more expensive than this months. So when they do the roll every month there's a slight drag. This is the problem with leveraged ETFs but also a number of popular non-leveraged stocks like USO and UNG.
I'm curious why you think there will be an increase in upcoming volatility intraday?
My opinion as a trader is: markets are random, scalping/daytrading is a hard (some would say impossible) way to make a living. I like to trade options because there are more ways to win. I can collect theta (time) decay even if the price stays fixed.
If for example you think bond yields will increase, then you're bearish on the bond market, so that view could be expressed by buying a calendar spread... eg in $TLT, sell the May $130 put and buy the July $130 put. The goal is that the May put decays in value while the long-dated put decays more slowly. Increased volatility will have an outsized effect on your long July option. You can do that trade for $192 per 1-lot.
But to be clear, that's very different than scalping in and out of a position intraday.
I sell premium as a primary investing strategy and as a rule there are only 20 or so underlyings with active enough derivatives markets by my standards. In widely traded underlyings not only do you get tighter markets, but a more valid crowd-sourcing of option probability.
As a rule, derivative liquidity in the bond market is significantly lower than equities.
What he is saying is that the Oct 15th move looks really strange, but not if you actually understand how fixed income market makers are reacting to the Volker rule. Quoting Dimon:
For instance, the total inventory of Treasuries readily available to market-makers today is $1.7 trillion, down from $2.7 trillion at its peak in 2007. Meanwhile, the Treasury market is $12.5 trillion; it was $4.4 trillion in 2007.
What he's saying is that market makers have smaller inventories on absolute terms, and much smaller inventories on relative terms, so when multiple big market participants try to move in the same direction at the same time, there is much less inventory available to soak up demand.