3 ms·

> Unfortunately the British financial regulators, who built London on a program of pretty much anything goes style regulation, woke up that day and decided that

by ethbro 11y ago

> Unfortunately the British financial regulators, who built London on a program of pretty much anything goes style regulation, woke up that day and decided that for the first time ever they needed to actually regulate and they declared Lehman's assets frozen, meaning that otherwise healthy funds (including pension, mutual, and hedge) now had no access to their own assets. I assume that there was reasoning behind this that's a little more nuanced, but even given that then I feel this was the main issue with the crisis. It seems like the financial system is continuously growing in complexity. In that while there were formerly N instruments, a year later there are N+z instruments. Coupled with the increased mobility of capital, and its tendency to seek the highest returns, you end up with large amounts of private money pooling fairly quickly in unexpected and not-broadly understood places (which is kind of self-evident: if everyone perfectly understood a particular instrument, you would not expect its use to outperform the market). And don't get me started on the continual improvement of ML systems: a closed loop where code constructs optimal instruments directly by parsing legislation and legal opinions no longer seems impossible. This seems like a problem in the event that something glitches from business as usual. You now have central bank officials (who while incredibly intelligent, by definition don't know as much about these specialized instruments) tasked with making quick decisions (because HFT and yay automated trading systems) without complete information. Summary: increased machine complexity and speed vs human decision as the only buffer to unintended emerging failure scenarios seems like a recipe for disaster

auntienomen 11y ago
You pivoted pretty quickly here from Lehman to HFT and automated trading. I think you're skipping over something very important by doing so: The mortgage backed securities and CDOs that Lehman and the other big banks got into trouble with are nearly bespoke contracts (more like a small town's bonds than an NYSE stock). Some fancy computing goes into determining what to sell any given contract for, but that's as far as it goes. They're traded by people who chat on the telephone, gather round the conference table, and then fax legalese back and forth. These contracts are not traded on the sorts of highly liquid public exchanges that have HFT. Nor is there much if any automated decision making in this corner of the financial industry. No point in paying nerds a fortune to get instant decisions when implementing said decisions is going to take days or weeks (if it's even possible). Which is to say: The 2008 financial crisis unfolded at human speeds. The public liquid markets didn't move any faster during the crisis than they in Herbert Hoover's day. (Investors already had telephones then.) Public markets react to data as it becomes available, and all the relevant data at the time was generated by humans. In particular, the central bank officials were forced to make quick decisions without complete information not 'because HFT and yay automated trading', but because there were a lot of well-informed and very scared humans telling them the economy was about to collapse.
- g10r 11y ago
  Absolutely.
- ethbro 11y ago
  Excellent distinction and point. Thanks for correcting me! Out of interest, are these bespoke because of the amounts we're talking about? In other words, I assume from your comment that institutional MBS and CDOs are a completely different beast than what I could buy from my broker, at least in size?
  - auntienomen 11y ago
    I said mortgage backed securities like mortgage CDOs are bespoke because the assets behind them aren't really fungible. Two CDOs might both derive their returns from home mortgages in the general Las Vegas area, but their return flows won't come from the same houses. This is very different from stocks, where any two shares in some company are equivalent (provided they're of the same share class). When pricing ABS, people try to pretend they are interchangeable, which is where the fancy computing and dodgy modelling comes in. My use of the word 'bespoke' here is somewhat poetic. CDO issuers are more likely to build a CDO and try sell it than they are to make them to custom specifications. (Investment banks did make custom CDOs & CDSs, of course. If an investor calls up and says "can you find X, and sell me some", you find X and sell him some. And yes, you can't just buy an MBS through your broker (unless you're incredibly rich). You can invest in mutual funds that own a wide variety of MBS, of course. But these things just aren't sold to retail investors. This is not so much because of the size of the assets involved -- you can slice off and sell as small a slice of the assets as you want -- but because the fixed costs of arranging a sale are high. You need to pay lawyers who know what the hell they're doing to set up the legal arrangements for these contracts. It doesn't make sense for Uncle Stocktrader to drop $$$$$ just so he can buy a $$$ slice of a CDO. Much easier to pool the costs in a mutual fund or a money market.
    - ethbro 11y ago
      Makes sense. I guess my central confusion is that I thought the entire reason for bundling mortgages into some kind of aggregate object (I'm going to drop the financial terminology, as I'd probably use it imprecisely) was so that they could transform it into a fungible asset and therefore have access to a broader market of customers who were interested in buying it. In other words, while retail investor might not be interested (either through a mutual fund or something else) in owning these 3 mortgages, they might be interested in owning a smaller share of 1,000 aggregated mortgages. And furthermore, that you could alter the criteria those 1,000 mortgages were selected by to create different products (offering different rates and risks). And furthermore, that you could attempt to value add (and I thought this was where it went off the rails) by packaging junk mortgages in some statistically magical way where you could sell the aggregate for more than the total value of its individual components. So, in essence my understanding was something like (start at top, trickle to bottom): Actual mortgage Basic mortgage backed security (no value add) Complex mortgage backed security (value add) Retail mutual funds / investment bank consumers (Which, as you've said, still has nothing to do with HFT, as given the nature of these things there probably isn't much pricing variance day-to-day, and therefore has little interest to that community)

by ethbro 11y ago

auntienomen 11y ago

You pivoted pretty quickly here from Lehman to HFT and automated trading. I think you're skipping over something very important by doing so: The mortgage backed securities and CDOs that Lehman and the other big banks got into trouble with are nearly bespoke contracts (more like a small town's bonds than an NYSE stock). Some fancy computing goes into determining what to sell any given contract for, but that's as far as it goes. They're traded by people who chat on the telephone, gather round the conference table, and then fax legalese back and forth. These contracts are not traded on the sorts of highly liquid public exchanges that have HFT. Nor is there much if any automated decision making in this corner of the financial industry. No point in paying nerds a fortune to get instant decisions when implementing said decisions is going to take days or weeks (if it's even possible). Which is to say: The 2008 financial crisis unfolded at human speeds. The public liquid markets didn't move any faster during the crisis than they in Herbert Hoover's day. (Investors already had telephones then.) Public markets react to data as it becomes available, and all the relevant data at the time was generated by humans. In particular, the central bank officials were forced to make quick decisions without complete information not 'because HFT and yay automated trading', but because there were a lot of well-informed and very scared humans telling them the economy was about to collapse.

g10r 11y ago
Absolutely.
ethbro 11y ago
Excellent distinction and point. Thanks for correcting me! Out of interest, are these bespoke because of the amounts we're talking about? In other words, I assume from your comment that institutional MBS and CDOs are a completely different beast than what I could buy from my broker, at least in size?
- auntienomen 11y ago
  I said mortgage backed securities like mortgage CDOs are bespoke because the assets behind them aren't really fungible. Two CDOs might both derive their returns from home mortgages in the general Las Vegas area, but their return flows won't come from the same houses. This is very different from stocks, where any two shares in some company are equivalent (provided they're of the same share class). When pricing ABS, people try to pretend they are interchangeable, which is where the fancy computing and dodgy modelling comes in. My use of the word 'bespoke' here is somewhat poetic. CDO issuers are more likely to build a CDO and try sell it than they are to make them to custom specifications. (Investment banks did make custom CDOs & CDSs, of course. If an investor calls up and says "can you find X, and sell me some", you find X and sell him some. And yes, you can't just buy an MBS through your broker (unless you're incredibly rich). You can invest in mutual funds that own a wide variety of MBS, of course. But these things just aren't sold to retail investors. This is not so much because of the size of the assets involved -- you can slice off and sell as small a slice of the assets as you want -- but because the fixed costs of arranging a sale are high. You need to pay lawyers who know what the hell they're doing to set up the legal arrangements for these contracts. It doesn't make sense for Uncle Stocktrader to drop $$$$$ just so he can buy a $$$ slice of a CDO. Much easier to pool the costs in a mutual fund or a money market.
  - ethbro 11y ago
    Makes sense. I guess my central confusion is that I thought the entire reason for bundling mortgages into some kind of aggregate object (I'm going to drop the financial terminology, as I'd probably use it imprecisely) was so that they could transform it into a fungible asset and therefore have access to a broader market of customers who were interested in buying it. In other words, while retail investor might not be interested (either through a mutual fund or something else) in owning these 3 mortgages, they might be interested in owning a smaller share of 1,000 aggregated mortgages. And furthermore, that you could alter the criteria those 1,000 mortgages were selected by to create different products (offering different rates and risks). And furthermore, that you could attempt to value add (and I thought this was where it went off the rails) by packaging junk mortgages in some statistically magical way where you could sell the aggregate for more than the total value of its individual components. So, in essence my understanding was something like (start at top, trickle to bottom): Actual mortgage Basic mortgage backed security (no value add) Complex mortgage backed security (value add) Retail mutual funds / investment bank consumers (Which, as you've said, still has nothing to do with HFT, as given the nature of these things there probably isn't much pricing variance day-to-day, and therefore has little interest to that community)