Arnold Kling  

Notes on McArdle's Law

Singapore: Where Do I Start?... Singapore's Policy Secret: Eco...

The theory that in a financial panic people demand too much risk protection is one possible explanation of what I like to call McArdle's Law, namely "Money is weird. Finance is weird."

In contrast, I quoted Hall and Woodward to the effect that mortgage securitization does not create supply chain problems that are any different from the supply chain problems faced by Toyota. Thus, they deny the validity of McArdle's Law.

This is an issue worth discussing at length.

One (minor) difference between Freddie Mac and Toyota is that Toyota's suppliers can't run and hide. If you manufacture paint or steel or electrical components, you have a fixed address and a lot of capital invested, so that if you fail to perform Toyota can hurt you. If you're a fly-by-night mortgage broker, you can dump a lot of bad loans into the system, take your profits, and by the time the loans default you could be in Tahiti.

Freddie Mac can manage the problem, but I think it's more costly to manage a supply chain where the bottom layer is a bunch of small mortgage brokers than to manage a manufacturing supply chain. If Toyota's paint supplier in turn is supplied by a firm that messes things up, presumably the problem with the paint will be pretty obvious to Toyota right away. If Countrywide Funding starts delivering loans to Freddie Mac from sleazebag mortgage brokers, it could be a year or more before Freddie Mac figures out that they have a problem. The process of identifying a sleazebag and fixing the problem can be quite expensive relative to the volume of loans involved. In the Coaseian model of the firm, I think that the case for vertical integration may be stronger with mortgage origination than it is with automobile manufacturing.

More importantly, with manufacturing, when the supplier delivers a component you can tell right then and there whether the component meets specifications. Once you accept delivery, you have no further need to require that your supplier have capital or put up collateral. If the seller goes out of business, you may lose a valued supplier, but the stuff you bought can still be used to make your widgets.

With financial contracts, such as bank deposits, mortgages, bonds, bond insurance, or swaps, you care about the other party's long term ability to meet its obligations. When you buy a bond, you can't inspect it, find that it meets specs, and say, "Phew! Now I don't care whether the company I bought it from goes down the tubes. " Quite the contrary.

What I am suggesting is that there is a fundamental difference between finance and manufacturing in terms of the role played by trust. In manufacturing, the length of time that a purchaser is exposed to risk from nonperformance by a supplier is relatively short. (When the length of time is longer than a few months, we often see warranties or performance bonds. For example, if company A relies on proprietary software from company B, company A may require that company B provide a copy of its source code in escrow, so that if company B goes bankrupt company A can take over the maintenance.)

With finance, trust is fundamental. Trust is based largely on financial practices that evolve over time. However, there is also a psychological component. When psychology changes, risk premiums change. Hence, McArdle's Law.

Comments and Sharing

COMMENTS (9 to date)
Tushar writes:

So Arnold

Considering the above, would you say this problem cannot be solved privately (assuming level-playing field regulations) ?

MattYoung writes:

We should consider a more sinister theory, something we can call reverse causality.

Freddie Mac aggregates loans on a national scale, way more aggregation than is needed for risk pooling according to the law of large numbers. They go beyond the economies of scale.

Loan brokers, in this scenario, might naturally raise the risk premium to bring the large of large numbers back into play.

Thomas DeMeo writes:

Why does trust need to be so fundamental to finance? I don't think it does. The mortgage industry could be organized into supply chains that work more efficiently and with far less need for trust.

ed writes:

"it could be a year or more before Freddie Mac figures out that they have a problem."

I don't understand this. Why can't you set it up so that the buyer (Freddie) takes a random sample of the loans and has someone go look at them more closely? (Sort of like you would do quality control in manufacturing.) If you did that, wouldn't the problems become apparent immediately? Even if there are tens of thousands of loans in the portfolio, you should only need to look at maybe 50 or so to get a pretty good idea what you've got. It wouldn't be free, but it should be worth if if there are hundreds of millions of dollars on the line.

I've been wondering this for a while. Did anybody do anything like that? Why not? What am I missing here?

Arnold Kling writes:


In 1990, I was the director of quality control sampling and reporting at Freddie Mac. Sampling is a good idea, but the process of re-underwriting loans is expensive.

It can work, but overall it is more expensive to inspect quality in than to design a better process (see Edwards Deming). A better process is old-fashioned mortgage lending, with the bank controlling the compensation and training of the people who originate its loans.

ed writes:

Thanks, Arnold.

I'm still a bit confused, though. Sampling and checking is expensive, but it seems mandatory in this case. Why would anyone buy a MBS if it hadn't been checked for quality by someone besides the originator? The stories I read about widespread sketchy practices and outright fraud lead me to believe that billions of dollars were committed with little or no quality controls at all. Were the checks done poorly, or was there no quality control system in place at all, or what?

I suppose this is a case where there were a few years where it didn't pay to be cautious, so all the cautious people got driven out of the market. Maybe situations like that don't arise so much in manufacturing.

Arnold Kling writes:

Ed,I'm not sure what happened, but my guess is that you have got it right--some people got into the market who decided that they didn't need quality control systems. As long as house prices are rising, you can do without that stuff.

And, as you say, it doesn't work that way in manufacturing. A car company can't get its bad workmanship overlooked because of some sort of price bubble.

Matthew C. writes:


Your recent posts on the financial crisis are very thoughtful and thought-provoking.

I'm not sure why, but you seem to have the best nose for practical economics of all the GMU bloggers. Perhaps it is your experience in the real world instead of just academia.

It is quite fascinating why companies purchased these securities. Of course, they were "rated" AAA. And many of the people selling them privately joked about what complete garbage they were.

I would suggest that there is another kind of bubble underlying many of the other bubbles we have seen recently. It is a bubble of CYA proceduralism. Rather than relying on judgment and intelligence, so many of our social systems increasingly rely on judgment via bureaucratic proceduralism. So if a bond is rated "AAA" then purchasing it must be OK. If a software program is "validated" with the appropriate poundage of documentation, then the system is necessarily OK, even if the users can tell you that it is a steaming pile of horse droppings.

Tom Grey writes:

The quality control function failure was in the ratings companies. AAA rated loan slices were NOT worthy of the rating -- but also, like Arnold notes, the unworthiness of the mortgages was not costless to check. And didn't matter in a (20+ years? from '82) rising market.

The risk model data probably didn't include any national housing price collapse, because there hadn't been any for so long; but the acceptance of crappy loans is the unleveraged heart of the sub-prime crisis.

Mathew, great insight about CYA proceduralism -- which is also related to academic credentialism.
And, in fact, the entire Ivy League (including Stanford) elite schools seem to be generating a class of elite bureacrats who want to know what procedures to follow so as to suffer no accountability negatives if something goes bad.

Comments for this entry have been closed
Return to top