Banking has suffered a not-sufficiently-acknowledged loss of know-how on the lending side. Back in the stone ages when I got an MBA, there were a few fellow students from big commercial banks like Chase (and no former investment banking analysts, the two year posts for college graduates) and they had all gone through credit training. But by the mid 1980s, the big end of town was fascinated with automated models and expert systems. American Express was considered the sophisticate of the crowd, allegedly with the most finely tuned program.
Fast forward 20 plus years, and the credit card issuers formerly seen as most savvy, Amex and Capital One, now are sporting credit losses not markedly different than their peers. FICO based mortgage lending has proven to be a train wreck.
The problem is that there isn’t a good substitute for knowledge of the borrower and his community. Does he understand what he is getting into? How stable is his employer? What are the prospects for the local economy? Those are important considerations, and they require judgment. That may still in the end be used as an input to a more structured decision process. but overly automating borrower assessment has resulted in information loss. It’s hardly a surprise that the quality of decisions deteriorated.
Meredith Whitney has pointed to this issue, but it has received surprisingly little attention:
Since the early 1990s, key bank products, mortgages and credit card lending were rapidly consolidated nationally. Banking went from “knowing your customer” or local lending, to relying on what have proven to be unreliable FICO credit scores and centralised underwriting. The government should now motivate local lenders (many of which have clean balance sheets) to re-widen their product offering to include credit cards and encourage the mega banks to provide servicing and processing facilities to banks that sold off these capabilities years ago.
Getting better customer input is crucial whether the powers that be are successful in putting in the reforms needed for private securitization to revive (the inaction on this front speaks volumes) or whether the end game is more on balance sheet lending by banks. The Financial Times’ John Dizard in April 2008 said that the Fed expected to see a considerable reversion to more credit intermediation by banks, but wasn”t taking the commensurate steps:
Think of the main US banks and dealers, along with their regulators, as the Iraqi government – though without the same unity, purpose or long-term planning. The cash positions and liquidity of both are better now. The Iraqi government is not squandering its money on food for the ration system, medicine, electric plant or water treatment.
The US banks and dealers are through the first quarter, and are backstopped by a Federal Reserve that has gone from vestal virgin to camp follower. Some of the accountants would have appended the above quote from Matthew’s gospel to their opinions on the banks’ and brokers’ quarterly earnings statement, but it did not fit the guidelines of SFAS 157, the accounting standard.
It is not fair to say the Fed does not have a plan. It does. The plan is for the banking system to recapitalise for a new on-balance sheet world by raising a minimum of $200bn in a short period of time, not longer than two quarters. That way, there is no credit crunch, according to the model.
We seem to be on to Plan B, which is to have the Fed step in to pretty much every credit market (adding commercial mortgage securities to the TALF is the latest wrinkle). If there ever is an exit, it would involve either a bigger role for traditional banks or considerable fixes to the securitization model, but we aren’t hearing any noise on either front.
I’d be curious if readers can point to milestones in this devolution. Some have suggested that the consumer lending skill loss took hold in the 2003 period onward, but Whitney pegs it as a 1990s phenomenon, and I saw some elements of behavior change even earlier than that. Any input here would be very useful.
13 comments
Comments are closed.
The 90’s or 80s were when tick box questionaires for loans started, so that is when the first signs of proper risk management decline arose.
What worries me is that the new Basel II rules are all about modelling risk and actively disencourage proper risk management.
I think much of it has to do with Insta-Credit. When the credit issuers started providing a lend/no lend decision in fewer than 5 minutes the FICO score became pre-eminant. At first this worked great because there was a lack of trust in the modeling and the only persons that qualified have very high FICO’s and the rest were sent for manual review. Even that is often done though without the lender so much as speaking to the customer. Of course this doesn’t apply to mortgages (at least not here in Canada). Credit card lending though is more about preventing fraud as long as the minimum FICO cut-offs are met, then digging into any of the 4C’s of basic lending. The computer does ALL that because the decision needs to be made quick. When I was a manager at one of these bailed out banks large processing offices we saw only 20% of applications for manual review. The computer did the rest of the instant gratification. Worse yet I actively worked with the credit bureaus on coming up with a similar system for small businesses, so they too could borrow without an owner cosigning….I’ve been out of the industry for 3 years now, but I wonder what that loan book looks like now.
YS:
When I graduated from Chicago, 1974, the most frequent employer of Chicago MBAs was commercial banks. The most common position, commercial lender. Other common jobs and employers: production line foreman (really, MBAs worked in plants producing goods), oil companies, CPA firms as auditors or consultants, management consulting firms and non-bank finance houses. I estimate only 10-12% of the class went to non-bank finance houses: investment banks, venture capital firms and the like.
In following MBA employment, I place the change in credit analysis in the 1980s too.
It was tough going to MBA school in the 1970s, you had to carry your broadsword at all times and lop off dinosaurs heads on the way to class!
I certainly share your concerns about the parallel phenomena of erosion of credit skills and excessive reliance on mechanical, quantiative techniques. The quantitative should have been used as terrific tools to enhance, not replace, credit skills. Maybe the MBA programs can help to start restoring some balance.
But I don’t think it’s only in the credit departments that we find folks failing to do old-fashioned homework. I’ve been appalled by the fund managers who failed to do the most rudimentary due diligence on Madoff-type schemes, such as “where are the custody controls”. And it’s not just fraud. Take, frex, the big-time pension managers who got caught not understanding the junior tranches they were buying in CDOs. But they could do a lot of handwaving with detailed graphs and charts about performance benchmarking. We can point fingers at a huge amount of misselling on the sell side, but it takes a lot of credulous supposedly-sophisticated stuffees on the buy side to produce these catastrophes.
So as I see it, it’s a broader erosion of responsibility for exercising good judgment that’s been accompanied by the increased use of quantitative CYA techniques. I don’t know how we go about changing incentives so that what was a “safe harbor” — see, I met the quantitative rules — doesn’t become a substitute for the exercise of responsible judgment.
I want, however, to come back to a theme from prior threads — your concern about whether/how to restart securitization and the future of dis- or re-intermediation.
You say:
Getting better customer input is crucial whether the powers that be are successful in putting in the reforms needed for private securitization to revive (the inaction on this front speaks volumes) or whether the end game is more on balance sheet lending by banks. So here are some questions.
What “reforms” do you think are needed to restart private securitization? Should those “reforms” put limits on securitization that would encourage re-intermediation, and if so, how would those “reforms” be designed in order to limit undue securitization risks? If so, how do you think we should be strengthening bank balance sheets so they will be able to handle a big chunk of debt financing moving out of the capital markets and back onto bank balance sheets? Re-intermediating is going to mean that the banks raise even more of their funding from the markets rather than from deposits. Which will make the “too-big-to-fail” problem potentially even larger.
I seem to be clearly missing something, but I’m not sure what we need in the way of “reforms” to restart plain vanilla securitization in housing and receivables. The impediments right now have to do primarily with risk appetite and pricing problems in the credit markets that aren’t unique to securitized assets. And at least for awhile, I doubt that “exotics”, where the disasterous risks were buried, are going to be marketable. So at least for awhile, we don’t need “reforms” to protect us from the high risk junk.
FWIW, Lew Ranieri (for whom I have a good deal of respect) thinks the steps the government is taking is making it possible for the plain vanilla stuff to start to get some traction. He said at the Miliken Conference that it will take awhile, but he’s surprisingly upbeat. And he’s not somebody whose simply a mindless cheerleader. He was certainly among those who were early in warning about the use of mortgage-backed markets to produce opaque layers of unmanageable risks. So though he’s clearly a supporter of securitization, he’s also well aware of its dangers. And he thinks the plain vanilla markets can recover.
Your thoughts?
I used to do a lot of work in AI and expert systems. I think the problem is people want to replace knowledge with automation, which is impossible. You always need the expert, the job of the system should be to make their work easier and not to replace them.
We’ve gone from a society that values knowledge, learning and effectiveness to one that values money, efficiency and speed. This inevitably leads to disaster in whatever form it takes.
I think that the deterioration in lending expertise is another aspect of the paper economy and tracks with its growth. That is to say that it goes back to the Reagan Adminstration. I would think that the Savings and Loan debacle would be an important milestone. Another would be Greenspan’s easy credit policies. When you had so much money entering the system, there was little incentive to monitor individual loans. Then you hit the Bush years where all these tendencies get magnified through deregulation, overleveraging, and bubble/Ponzi economics.
Re: milestones in this devolution >>
The 1978 Supreme Court decision Marquette National Bank v. First of Omaha Service Corp. concluded that national banks, such as Bank of America and Citibank, can charge the highest interest rate allowed in the bank's home state, regardless of where the borrower lives. This means that credit card issuers located in states with liberal or nonexistent usury laws, such as Delaware and South Dakota, can "export" the lack of an interest rate cap to customers in states with usury laws in place. These companies can ignore the "Natural State's" 17-percent constitutional limit.
See and hear:
Thomas Geoghegan on “Infinite Debt: How Unlimited Interest Rates Destroyed the Economy”See and enjoy also:
You Never Give Me Your Money(long 6 min version)-The Beatles
The problem is that bank officers have outsourced risk – it used to be that banks actually needed to know and be apprised of the client. Now that risk is securitized and packaged into someone else’s problem (aka supposedly the mkt). That explains why the skillset of compliance and risk went out the door.
B/c supposedly OTC mkts are excellent places for price discovery.
It is now profoundly obvious that the stress test was an ill-conceived idea and a misnomer. The “all is well” tests would have been a better name.
Dean.
I think nazhedzha’s post is one of the better posts we ever see here on NC, but as much as I respect Ranieri, I’m not sure the spreads in vanilla securitizations could ever support the transactions costs involved (without duplicity or massive fraud by rating agencies and the likes of AIGFP). Why would anyone expend the effort to engage in securitizations (with all the legal and finance costs involved) without a payout that is above market? Color me skeptical.
Automated vs. credit judgment/skill is missing the point.
I’ve worked in risk management for 10 years at top banks and never seen a expert judgment manual system that outperforms a properly designed and executed automated/scored system.
The problem is: idiots without an ounce of common sense built the automated systems. They basically looked at credit bureau data, but ignored such trivia as borrower’s ability to pay or collateral quality. And the quality of appraisals and screening by loan officers, oh, nevermind.
I would argue they built it exactly how they and their management wanted it built. Shocking – they did exactly what they were paid to do.
As long as you give people incentive to just grow next quarter’s earnings, damn the long term, you’ll get the same result.
But focusing on the degree of automation or judgment – that’s at best a symptom, and in my view, not ever a true underlying cause of the problem.
@Anonymous Jones: Thanks for your comment! Let me follow up on your skepticism, however. You wrote:
Why would anyone expend the effort to engage in securitizations (with all the legal and finance costs involved) without a payout that is above market? Color me skeptical. Because the plain vanilla stuff involves neither huge transaction costs nor an above-market payout.
Basic securitization, involving standardized, well-understood cash-flow streams with a pretty big equity cushion embedded in the structure was a very valuable financial innovation. It uses the capital markets to match risk/yield/duration preferences of different classes of investors. Take conforming mortgages with max 80% LTV ratios and what used to be pretty standard credit assessments of repayment capacity of borrowers. Standardization and a sizeable equity cushion means that the transaction costs of bundling, tranching and marketing the securitized assets are modest relative to the size of each bundle. The cash flows from the bundles could be sliced to give different risk/yield outcomes. The investors who bought the safest tranches were paying more (getting lower yield) than the mortgage market for individual loans. The yield differential went to investors who were looking for a higher yield and willing and able to take on greater risk.
What destroyed securitization was adding a whole bunch of bells and whistles to try to get to the same result as plain vanilla securitization. So they used models that relied heavily on diversification, ignoring that the apparently diversified bundles would heavily correlate in wide-spread economic distress. (Plalin vanilla securitization does include diversification as one technique to enhance risk protection but doesn’t rely primarily on diversification.) They added wraps from AAA rated institutions where neither the rating agencies nor the markets were taking into account the extent to which the AAA status was being destroyed by the promiscious monetization of the AAA rating. They added to homogenous bundles of mortgages lots of odds and ends of “highly rated” junk when even their poorly designed models didn’t produce the right results. They layered securitized assets and derivatives in increasingly exotic and model-dependent instruments that appeared like magic to make risk disappear.
It’s this stuff that ultimately doesn’t make sense financially due to transaction costs and “an above-market payout”. It’s not going to come back — at least for a few years.
But the inefficiences of exotics doesn’t change the cost-effectiveness and market efficiences (intermediating borrower and investor preferences) of plain vanilla securitization.
In general, I agree with lanhamac. Automated underwriting of individual credit risks can work effectively. However, when FICO became the single most important metric, and when everyone knew it — there had to be some problems.
When Suze Orman started talking about it, and people explicitly worked to improve their scores, their historical predictive value had to decrease.
At least some people were clearly gaming the system.
However, the idea that you could do low doc, then no doc mortgages for hundreds of thousands of dollars is inherently goofy. I think people knew that they were relying on the value of collateral — housing values — to support the loans. When they tanked… we all know the story.
Maybe even more important was the disintermediation of traditional lenders and the idea that underwriters with no financial incentive to make sound loans but enormous incentives to make a lot of loans were doing the bulk of mortgage lending. It was a rush to the bottom.
Who knows what old fashioned, conservative, traditional bankers would have done if they could have just sold the loans.
The incentives were all screwed up….