Hello,There is a new paper. Technical comments are welcome. Thank you in advance.New PaperNassim

I looked briefly. The Jan 21, 2008 market event is interesting and there are various similarities/differenceswith the Oct, '87 crash, which you might discuss. In both cases there was a 'forced selling' triggered by abouta $80 bil position. The '87 crash was triggered primarily by the front-running of the known intentionsof the Leland, O'Brien, Rubinstein firm (LOR), and it turned out that LOR didn't sell that much.Since your point is that ten $8 bil sales would not be so disruptive, it suggests that large institutions in the marketneed to be forced to think through their order management practices prior to a panic event.

Stable distributions make no difference. Solved for ALL unimodal distributions.

Interesting work! There's a key example of a large centralized system that you might want to add to your list.Systems in which the components are coupled by global markets and correlated asset holdings show the same diseconomies that speak about. For example, even if all banks are small, to the extent that they share identical price data on the same pool of assets (e.g., mortgages), then the system is a large fragile one in which the failure of one small bank increases the probability of failure of other small banks due to the impact of one bank failure on asset prices which, in turn, impact the balance sheets of all other banks holding the similar assets. The problem may be further exacerbated if all banks, are subject to similar time-varying reserve requirements.

QuoteSystems in which the components are coupled by global markets and correlated asset holdings show the same diseconomies that speak about. That's just about everything then? We then have one 'chain of faulure'.Can't see an alternative. Maybe I've missed something.

QuoteOriginally posted by: CuchulainnQuoteSystems in which the components are coupled by global markets and correlated asset holdings show the same diseconomies that speak about. That's just about everything then? We then have one 'chain of failure'.Can't see an alternative. Maybe I've missed something.At some level, you are right that there is no easy alternative. The world has inexorably moved from the small (largely independent economies and independent financial systems in respective countries) to the large (a global economy with a global financial system). This problem is well known to epidemiologists and ecologists who fear that the rise of global travel and commerce mean that contagion can race unchecked around the world. By that analogy, the only solution is a series of quarantine zones that isolate chunks of the financial system from each other. I doubt that can happen.But that doesn't mean we can't create incentives for some beneficial logical separations within the globe's portfolio of financial institutions. If we make each financial institution's tax rate proportional to that institution's correlation to other institutions, we create a direct disincentive for herd-like investment in bubbles. A tax on correlation would encourage each institution to find uncorrelated or anticorrelated investments rather than join the bandwagon investment du jour. A correlation tax is a strong incentive for genetic diversity in financial institutions and reduces the chance of some particular investment strategy becoming too big to fail.

Any time there are speculators - any time an asset's sale can be postponed - it creates problems for a price signal. Because you cannot separate the demand of end users, and the supply of producers, from the demand of speculators in the price. And obviously speculators can change their plans much more quickly than producers or users.A price may not be an "equilibrium" price, but simply a popular and temporary guess. It can help if you know that, for example, 50% of homes are owned by owners, and 50% of homes are owned or being built by speculators, whose intention is to sell them in the near future, rather than live in them. But this sort of information is impractical, and most economic activity evolves around the price ticker.There is a simple solution: price instability. In currencies, for example, trend traders push the prices as far as they dare one way and the other, so that immediate and long-term volatility generally match. So more specifically, the solution is immediate price instability that is indicative of long-term instability.The immediate price instability tends to regulate the balance of speculators. It is not random that housing, which is very illiquid and hard to change the location of, is where we have seen the most problems lately. Rather than in oil, or something, where people think of there being a lot of speculators. Because trend traders can run the price of oil up and down until they shake off speculators and run their stops, and at the same time find the bids and offers of end producers and consumers.So price instability in a liquid market tends to weaken the signal of speculators, which can lead to redundant inventories. And it reveals the signal of actual supply and demand levels of end users. The most useful piece of information would be like actual curves and elasticities for supply and demand. Running the price up and down the chart - such that speculators cannot hold prices at a single level long enough to completely hide actual demand levels - reveals such curves sort of like a sweeping radar image reveals 360 degrees of airspace.If you wanted to perform an economic experiment to determine actual demand for any good, what would you do? You would try offering it at different prices for periods of time, and see the quantities demanded at each price. Running it up and down extremely fast would allow you to detect a trend, like a really fast spinning radar gives you enough sample points to plot motion in a single craft.One implication of this, is that a "flash crash" is an extremely useful tool, like a fire drill. The more flash crashes there are, the healthier a market will be so far as planning patterns of production. We didn't have any flash crashes in housing for many years, and it certainly was no help to planning optimal leverage and patterns of production and speculation. A speculator must plan for a range of outcomes, and his plans are better informed by observing a range of outcomes while making his plans.

QuoteOriginally posted by: farmerAny time there are speculators - any time an asset's sale can be postponed - it creates problems for a price signal. Because you cannot separate the demand of end users, and the supply of producers, from the demand of speculators in the price. And obviously speculators can change their plans much more quickly than producers or users.A price may not be an "equilibrium" price, but simply a popular and temporary guess. It can help if you know that, for example, 50% of homes are owned by owners, and 50% of homes are owned or being built by speculators, whose intention is to sell them in the near future, rather than live in them. But this sort of information is impractical, and most economic activity evolves around the price ticker.There is a simple solution: price instability. In currencies, for example, trend traders push the prices as far as they dare one way and the other, so that immediate and long-term volatility generally match. So more specifically, the solution is immediate price instability that is indicative of long-term instability.The immediate price instability tends to regulate the balance of speculators. It is not random that housing, which is very illiquid and hard to change the location of, is where we have seen the most problems lately. Rather than in oil, or something, where people think of there being a lot of speculators. Because trend traders can run the price of oil up and down until they shake off speculators and run their stops, and at the same time find the bids and offers of end producers and consumers.So price instability in a liquid market tends to weaken the signal of speculators, which can lead to redundant inventories. And it reveals the signal of actual supply and demand levels of end users. The most useful piece of information would be like actual curves and elasticities for supply and demand. Running the price up and down the chart - such that speculators cannot hold prices at a single level long enough to completely hide actual demand levels - reveals such curves sort of like a sweeping radar image reveals 360 degrees of airspace.If you wanted to perform an economic experiment to determine actual demand for any good, what would you do? You would try offering it at different prices for periods of time, and see the quantities demanded at each price. Running it up and down extremely fast would allow you to detect a trend, like a really fast spinning radar gives you enough sample points to plot motion in a single craft.One implication of this, is that a "flash crash" is an extremely useful tool, like a fire drill. The more flash crashes there are, the healthier a market will be so far as planning patterns of production. We didn't have any flash crashes in housing for many years, and it certainly was no help to planning optimal leverage and patterns of production and speculation. A speculator must plan for a range of outcomes, and his plans are better informed by observing a range of outcomes while making his plans.Actually, price instability fails to reveal supply and demand because the tactic has two key assumptions. First, it assumes that the volume of purchasing of a product is equivalent to the volume of demand for the product. If a grocery store cuts the price of diapers in half, it will sell a ton of diapers in a short period of time but that's not because parents decided they can put their babies on high-fiber diets. Instead, the parents are speculating that the price of diapers will revert and they are increasing their household inventories of diapers in advance of the price increase when the grocer ends the sale. Second, the tactic assumes that genuine producers and users are risk-neutral. If a potential cookie factory owner sees volatile prices for flour, sugar, food oils, etc., then they might decide not to build a new cookie factory because the chance of the volatility accumulating to high ingredient prices (making cookies too expensive) would be too high.Small price excursions do strengthen those exposed to them but only if training matches reality and only if downstream players understand what's happening. If primary participants learn that someone is injecting 1% to 5% flashcrashes as artificial training, they will soon adapt their systems as if a 5.001% event cannot happen. And if secondary participants aren't aware that the primary participants are buffering the volatility (so that secondary players don't see any of the 1%-5% blips), then any shock bigger than 5% will wreck havoc on everyone.

One of the issues that has been exercising my mind with regard to the markets current love affair with risk parity is the degree to which there may already be be crowding into the same risk factor trades due to all of the necessary dimensional reduction via PCA etc. This may have the effect of increasing the overall systemic risk along the lines of the following paper

Doesn't PCA fall into the family of tools based on L2 norm, and thus according to the book that is the subject of this thread (p60): "fundamentally wrong...and not dependable enough to be of use in anything remotely related to risky decisions" ?Do we know when the final version of the book will be published?

Indeed PCAs will be necessarily unstable ?anything related to variance.So are correlations of course.

The central problem of financial risk is not mathematics, it is topology. Suppose there is a gas station right on the highway, one a few blocks off the exit, and one a mile out in the wilderness. Their prices will be predictable. Now suppose there is a change of topology. The remote gas station might buy a billboard. A new app or robot car might be programmed to drive to the cheapest gas. People may start to drive flying cars. The gas stations may become redundant. One gas station may go out of business. The two remaining stations may specialize, so as to not be redundant.The stability of traditional businesses is driven by geographic and other types of friction. I theorized a few years ago that demand for option-like products should decline, as businesses adapt to a new Internet-oriented and global topology. In the early 1980's, the introduction of S&P 500 futures changed the market topology, and led to a crash. Similarly, a leveling of the playing field between floor traders or locals, and Globex traders all across the globe, created a chance for redundancy. New hedge funds started to participate in new markets. New frontrunners emerged to trade ahead of their orders. The floor traders found other opportunities.During the dotcom boom, John Chambers of Cisco popularized the term "visibility" regarding his ability to optimize production and inventory for random events coming up the supply chain. Is it possible that, having adapted to global electronic trading and Internet information, we have entered a new area of stability? China is no longer exploding, Europe is no longer panicking, Batista is done growing. Changes is topology, like building a new Walmart in a rural area, cause much more painful changes than monetary policy, in my opinion.