Alas, the real answer is even longer then the long answer. In essence, it was the correlation and the counterparty risk which brought all that system down.
1) The correlation between different mortgages is extremely difficult to assess and even more difficult to hedge properly. And the situation becomes even more dire if you try to do that with different CDO tranches. Financial engineers knew it all along and were screaming that what was being done is wrong (see http://www.forbes.com/2008/10/07/securities-quants-models-oped-cx_ss_1008shreve.html). But nobody listened...
2) To be fair, many companies which went into trouble did have their positions hedged against the moves in the housing market (according to their models, of course, which were not precise, see 1) ); what they didn't take into account is the counterparty risk: most of the hedging on that market is done through swaps, so when the counterparty in the swap defaults, it leaves the company position "naked". Thus, they thought they are hedged, but they were not. At the wake of the crisis even the most sophisticated companies (think Sachs) couldn't evaluate their exposure to different companies: it is indeed difficult as they all hedge with each other....
BTW, the statement that "those models assumed that home values would continue to rise" is patently wrong: none of the mathematical models of credit derivatives assumes that. But discussing that will take a good lecture and I will need to embed formulae in lj to explain what is wrong with that statement -- let's spare our host the trouble of cleaning up the mailbox :)
One more thing: I don't believe in the irrational behavior in the complete market (i.e. herding, loss aversion, hyperbolic discounting, etc) -- simple undercutting argument would prove that it can't exist any prolonged period of time.
Re: Напомнило
1) The correlation between different mortgages is extremely difficult to assess and even more difficult to hedge properly. And the situation becomes even more dire if you try to do that with different CDO tranches. Financial engineers knew it all along and were screaming that what was being done is wrong (see http://www.forbes.com/2008/10/07/securities-quants-models-oped-cx_ss_1008shreve.html). But nobody listened...
2) To be fair, many companies which went into trouble did have their positions hedged against the moves in the housing market (according to their models, of course, which were not precise, see 1) ); what they didn't take into account is the counterparty risk: most of the hedging on that market is done through swaps, so when the counterparty in the swap defaults, it leaves the company position "naked". Thus, they thought they are hedged, but they were not. At the wake of the crisis even the most sophisticated companies (think Sachs) couldn't evaluate their exposure to different companies: it is indeed difficult as they all hedge with each other....
BTW, the statement that "those models assumed that home values would continue to rise" is patently wrong: none of the mathematical models of credit derivatives assumes that. But discussing that will take a good lecture and I will need to embed formulae in lj to explain what is wrong with that statement -- let's spare our host the trouble of cleaning up the mailbox :)
One more thing: I don't believe in the irrational behavior in the complete market (i.e. herding, loss aversion, hyperbolic discounting, etc) -- simple undercutting argument would prove that it can't exist any prolonged period of time.