During the breakneck expansion in consumer credit in the past five years, bank executives often argued that automated credit approval processes and computer modelling techniques had improved their lending criteria and would allow them to avoid the problems of the past.

However, the growth in bad debts has shown them they still have more to do.

Lloyds TSB said this week it had adjusted its underwriting models so it now looks at the propensity of customer to repay. It uses statistical models to identify the key characteristics of cohorts of good paying customers against those who are likely to default. Other banks are increasingly taking a similar approach.

These techniques have been well developed in the United States, where banks and specialised lenders employ highly qualified computer experts to build models to help spot people that may become bad credit risks. At the same time, British banks have taken the old-fashioned approach of simply saying “no” more often.

They have become more stringent about which customers they lend to and have started rejecting more loan applications.

Barclaycard, the country’s largest credit card operator, last year increased its rejection rate and it is turning down 55 per cent of new credit card applications – a much higher rate than in previous years.

This year, HBOS has reduced the credit limits for 600,000 credit card accounts so customers can borrow less.

Greater sharing of data by banks about their customers should also limit future problems. In the past, borrowers were only reported to credit ratings agencies after they defaulted on a loan.

Today, banks are able to access more information about other loans a consumer may have before making a decision, allowing them to weed out potential problems.

Copyright The Financial Times Limited 2018. All rights reserved.

Comments have not been enabled for this article.