The subprime crisis is often explained in terms of trust: one day trust between financial counterparties vanished and here was a crisis. Explanations in terms of “market confidence” refer in fact to two distinct phenomena, one being indeed trust and the other one being more plainly straightforward profitability. Let me start with profitability. Subprime loans had been repackaged in their thousands so as to mimic a straightforward obligation, be these Asset-Backed Securities (ABS) or – in a second step – collateralized Debt Obligations (CDO) comprising ABSs. When the risk-based premium embedded within the interest rate charged on a subprime mortgage revealed itself to be insufficient to cover the actual risk of default on such loans, investors stopped buying the products where these had been repackaged. It’s as simple as that and has of course nothing to do in this case with either trust or confidence: investors simply stopped purchasing a product that had ceased to be profitable.
As I’m aiming – as I’ve explained in earlier blogs (*) – at a “complete” explanation of the subprime crisis, I need to model this explicitly – and writing it as part of a computer program is a good manner for ensuring it is complete. Resorting then to some pseudo-code in the style of BASIC I would write something like the following:
If Profitability < 0 then “Don’t buy”
The zero in the equation can be replaced by any figure you wish, depending on your profitability target.
Trust slipped in under two different guises. I remind that a Collateralized Debt Obligation is a composite financial product the elements of which are certificates also called “tranches” of Asset-Backed Securities. Usually a hundred or more ABS tranches are thus repackaged as a CDO. Apart from being backed by subprime loans, ABS can alternatively be backed by credit card debt, auto loans, etc. The composition of a CDO is advertised in its prospectus, so it is possible to know whether or not a particular one comprised ABS backed by subprime loans. Some CDOs are more complex, like a “CDO square”, a CDO composed of CDOs, meaning that it is not unusual for a “CDO square” to be composed of over a thousand underlying ABSs. And this is where trust kicked in back in August 2007: if you couldn’t tell if a financial product contained or not subprime loans, you would simply abstain from buying it.
Confidence is a bit trickier than defining a profitability threshold under which you abstain from buying. You would need something like:
If some part of Product is subprime loans then “Don’t buy”
Typically you would have an array containing the list of components of your CDO and you would loop through it and if subprime is encountered that would be the end of it.
Deal = “Fine”
For each Part in Product
—If Part = “subprime” then
——Deal = “Not Fine”
——Exit For loop
—End If
Next
If Deal = “Fine” then “Buy”
If you were able to loop all the way through without being kicked out and Deal being redefined as “Not Fine” then the deal is fine and you can buy.
But it goes further than that as the condition should also comprise the case where it is impossible to establish whether or not the product contains subprime loans, say that the label for Part is missing making it impossible to find out if it is subprime or not. The line
If Part = “subprime” then
would be rewritten then in the following manner, taking into account the possible occurrence of a missing label, showing as an empty string:
If Part = (“subprime” OR “”) then
emphasizing that the components of the product need to be explicitly known for the deal to go through.
Now an extension of the trust element was also involved: counterparty risk. It might be that you were not buying a single product but engaging in a long term relationship with a counterparty. As with an interest rate swap, say, where payments are made every six months over a period of several years. It is crucial to know that your counterparty will be there in the long haul. Now how many subprime loans has your counterparty in its portfolio that may weaken it financially so that its solvency over the long term might be compromised? Difficult to know. Let’s turn again towards the simple-minded but enlightening process of translating the condition into a line of programming.
If Deal = “Fine” then [passed successfully through the loop above]
—If Number of Cash Flows > 1 then [Product is long term]
——If Counterparty has (No subprime loans) then “Buy”
—End If
End If
The last condition would also be rewritten as a loop
CounterpartyPortfolio = “Fine”
For each Part in Portfolio
—If Part = (“subprime” OR “”) then
——Count = Count + 1
——If Count > 100 then [100 or any other threshold]
———CounterpartyPortfolio = “Not Fine”
———Exit For loop
——End If
—End If
Next
If CounterpartyPortfolio = “Fine” then “Buy”
Bringing this all together, the whole “trust” or “confidence” issue would actually amount to the following test:
Deal = “Fine”
For each Part in Product
—If Part = (“subprime” OR “”) then
——Deal = “Not Fine”
——Exit For loop
—End If
Next
If Deal = “Fine” then
—If Number of Cash Flows > 1 then
——For each Part in CounterpartyPortfolio
———If Part = (“subprime” OR “”) then
————Count = Count + 1
————If Count > 100 then
—————Deal = “Not Fine”
—————Exit For loop
————End If
———End If
——Next
—End If
End If
If Deal = “Fine” then “Buy”
As can be seen from my pseudo-code, what the “trust” or “confidence” issue means in fact in the subprime crisis is that the conditions for purchasing a financial product become increasingly restrictive. The number of hurdles grows and the information necessary for a transaction to be allowed is growing in step: first about the product, containing subprime loans or not, about the seller next: “containing” (in portfolio) subprime loans or not. The crucial part is the “If Part = (“subprime” OR “”)” test: a large number of transactions would be prohibited because of the explicit presence of subprime loans but many more because of the missing information represented by the empty string “”: in case of doubt please abstain!
(*) The subprime crisis: a human complex system phenomenon, Agents using financial models and the “human cognitive cocktail”, Pricing models: why the good ones are useless and the true ones, priceless