Monthly articles (English and French) on the theme "Querying economic orthodoxy"

No. 40 - April 2009

Subprime productivity


Banking will have to switch from being a raucous, entrepreneurial, and aggressive industry into a dull, safe and conservative one.

Matthew Lynn, Bloomberg Company News (re Royal Bank of Scotland), 15 October 2008

The application of industrial-style productivity to financial services has got us into terrible trouble.

More with less effort

Productivity is a measure of the efficiency of an organisation, of how much it can achieve from the work of the people it employs. Ever since the end of the middle ages, industry, commerce and agriculture have been enhancing their productivity, finding ways of producing more with less human effort.

In the late nineteenth century, almost a fifth of the British working population still worked in agriculture; today the proportion is less than one fiftieth, yet Britain grows more food than in Victorian times.

In manufacturing, robotic production-line machinery has made possible huge productivity increases. In the late 1970s, the old British Leyland group, notorious for its chronic inefficiency, made about six cars per employee per annum. Today, the most efficient firms make more than 60 (1). We can be happy about this progress; it means that most of the unspeakably boring work of the production line is now done by robots, and generally done better.

Engineers now use machine tools that can automatically perform in sequence a wide variety of cutting and drilling operations, turning blocks of metal into intricately-worked products such as cylinder blocks, gearboxes or valve bodies. And they can do this repeatedly for hours on end without human intervention.

Another remarkable metamorphosis has been achieved in steelmaking. Here, steel slabs are transformed into rods, strips, girders, rails, etc. by the process of 'rolling'. This means passing hot steel between successive pairs of steel rollers (called simply rolls in the trade); each pair resembles a heavy-duty industrial version of the traditional washerwoman's mangle.

Vivid pictures of old-time steelmills show teams of brawny men manipulating slabs of white-hot steel with giant tongs, pushing them back and forth, back and forth through one set of rolls after another till they were reduced to their final shape. Such spectacular heroics belong to history. The modern rolling mill, in which steel passes continuously through a long series of rolls, is totally automated. So much so that Sir Ian MacGregor (2), chairman of British Steel in the 1980s, once observed that you should be able to fire a shotgun down the shop-floor without hitting anyone.

Why not automate banking too?

Can this kind of automation be applied to financial services? Many people today think this is perfectly possible. Or, perhaps we should say, they want to believe it possible; for, in finance as in manufacturing, automation offers the possibility of drastic cost-cutting. And if business is extremely competitive, as conventional economists want it to be, then the need to cut costs is ever-present and compelling.

Consider bank lending. Traditionally, the arrangement of large personal loans such as mortgages was a laborious process, for good reasons. When a bank customer asked for mortgage, it was thought necessary to examine the borrower's history and circumstances very carefully, so as to be reasonably certain that he or she would be able and willing to keep up the repayments and would not become a bad debtor. In Business Week, Dean Foust and Aaron Pressman describe (3) this time-honoured routine: banks would go to great lengths to vet potential borrowers, checking pay stubs and tax returns, calling employers, poring over investment account statements, and on and on, a process called underwriting.

More recently, however, this task has been largely automated by computer analysis of the data in loan applications, and by the use of 'credit scores'. In a recent IMF working paper (4), John Kiff and Paul Mills note that automated underwriting (using computer models rather than loan officer judgment) has made loan origination more efficient. They quote specialist studies which indicate that savings [estimated at] up to 3% of total loan value are associated with automated underwriting and that such savings may average $916 on each loan.

The online business dictionary Investopedia explains (5): Automated underwriting engines can provide near-instantaneous loan approval or denial decisions; therefore, implementing automated underwriting systems can save a considerable amount of time, as manual underwriting can take as long as 60 days to complete. In addition to the time savings, automated underwriting is preferred because it is based on algorithms, eliminating human bias.

The credit scoring shortcut

The credit score is a figure which is supposed to indicate a borrower's creditworthiness. In America, these scores have come to be very important in determining one's access to credit, whether to pay everyday bills by credit card, or to borrow on a bigger scale to buy a car or a house. The most widely-used score is the FICO, named after Fair Isaac Corporation, a US company set up in 1956 by engineer Bill Fair and mathematician Earl Isaac.

Scoring works by analysing a person's credit history. Every time one borrows, the lender passes details of the loan to one or more of the credit bureaux (Equifax, Experian, Transunion and others), which keep records of personal lending transactions and of repayments. So these agencies have detailed information on how much and how often a person has borrowed, and whether or not the borrower has a history of repaying on time.

Banks and other lenders have access to all this data, but it is voluminous and troublesome to handle. The bright idea of Mr Fair and Mr Isaac was to collect the data and analyse it themselves on their computers, thus producing a 'credit score' for just about every American adult. Their system is now used in many other countries. Such scores are widely used by lenders as a quick guide to the creditworthiness of those who want to borrow.

The credit score alone is far from adequate for deciding whether or not someone should be lent enough money to buy a house. The data on which credit scores are based are derived from past borrowings; they reveal nothing about the borrower's current income or other circumstances. Moreover, there are various devious means by which credit scores can be 'doctored' so that they look better than the real credit history justifies.

Careless lending

So further research is still necessary before a loan can reasonably be offered. But in the recent American housing boom, many lenders became extraordinarily careless. Real-estate lawyer Michael Frachioni writes (6) that, while carrying out an audit of a large lending institution in 2003, he found that appropriate documentation or information was either incomplete or missing entirely in over a quarter of all new account files.

In a climate of persistently rising house prices, it seemed for a while that mortgages would always be safe, since the value of the property backing them would keep on rising. It appeared superfluous to fuss over details such as whether a borrower actually had enough income to service the loan. If it turned out that he hadn't, the bank could always foreclose and sell the house at a profit.

Katalina Bianco of the publishing house CCH describes (7) 'stated income' loans, for which the borrower does not have to provide documentation to substantiate the income stated on the application to finance home purchases. She continues: in 2007, 40 percent of all subprime loans were generated by automated underwriting. Automated underwriting meant minimal documentation and much quicker decisions, sometimes as soon as within 30 seconds as opposed to the week it would take for an underwriter to generate a decision. An executive vice-president for Countrywide Home Loans (8) also noted in 2004 that 'previously, every mortgage required a standard set of full documentation.' Many experts believe that lax controls and a willingness to rely on shortcuts led to the approval of buyers that under a less-automated system would not have been approved.

An automation too far?

One has to wonder whether it really makes any sense to automate and vastly accelerate the 'production' of mortgages, as if they were cars or dishwashers. A mortgage is not a material product; it is a contract between a bank and an individual human borrower. So it is subject to the vagaries of human nature and the vicissitudes of human circumstances. Setting up such contracts without careful consideration is asking for trouble. The old, slow method of arranging mortgages gave lenders and borrowers time to ponder over whether they were doing the right thing. The hectic pace of the electronic age leads to too many mistakes. When these involve house purchase, the consequences can be tragic for the people concerned. When there are millions of such mistakes, they can be disastrous for the worldwide economy.

And there is another problem. 'Higher productivity' in banking is a euphemism for fewer of the good steady jobs that banks used to provide. No matter, say the free-marketeers, cheaper bank services mean that bank customers have more cash to spend on other things, thus creating more jobs in other sectors. Today, all that chatter rings pretty hollow. It would be a good idea to return to sound, careful, low productivity, high quality banking. To employing real bankers in bank branches, rather than a few technicians in remote data processing centres, out of touch with the customers and their particular needs and circumstances.

This would be more costly, less 'efficient', and slower than high-tech automated underwriting. But cheap, quick and nasty retail banking has been given a pretty thorough trial. Are we happy with the consequences?

* * * * *


1 Toyota and Chrysler led the six largest multi-plant North American automakers in total manufacturing productivity (assembly, stamping, engine and transmission), each averaged 30.37 labour hours to manufacture a vehicle...Reuters' report 5 June 2008,
see http://www.reuters.com/article/pressRelease/idUS150987+05-Jun-2008+PRN20080605

If we assume a 40-hour week and 48 working weeks in the year, then 30 labour hours per vehicle = 64 vehicles per person per year.

2 Sir Ian McGregor (born 1912) became chairman of British Steel Corporation in 1980 and carried out drastic restructuring of the business. It was said of him that McGregor will never go to hell; the Devil will not let him in, lest he shut down half the furnaces.

3 See Credit Scores: Not-so-Magic Numbers in Business Week, 7 February 2008

4 John Kiff and Paul Mills, Money for Nothing, IMF working paper 07188 (July 2007), page 3: see http://www.imf.org/external/pubs/ft/wp/2007/wp07188.pdf

5 See Michael I. Frachioni, Without a net in The RMA Journal (The Risk Management Association, Pittsburgh, Pennsylvania), June 2004

7 See http://www.bestcase.com/grafix/pdf/Subprime_08.pdf, pages 7 and 8. CCH, part of the Wolters Kluger group, is a leading US publisher specialising in tax and business law.

8 Countrywide, a leading California-based mortgage lender, ran into serious trouble as a result of its involvement in subprime lending. It was acquired by Bank of America in July 2008.