Misconceptions over AMA

Financial services firms have serious misconceptions about the value Basel II's advanced measurement approach (AMA) can bring to their organisation, according to the latest OR&C Intelligence survey.

The survey, sponsored by London-based consulting firm Chase Cooper, showed that firms who have completed implementing the AMA believe it is more beneficial than those who are planning to do the AMA. Of those respondents who said they'd implemented the AMA, 66.7% said it provided "significant" benefit to the business line, while 33.3% said it provided "some benefit" to business lines. None of the respondents said it provided small or no benefit.

In contrast, among those who said they planned to use the AMA, only 37.5% said it would give their firms "significant" benefit, while 47.9% said it would offer "some" benefit. Another 12.5% claimed it would offer "small" benefit while another 2.1% claimed it would provide no benefit at all.

"There are two ideas that have come together. On one hand, organisations have got the idea that AMA is very complicated and very expensive and requires vast amounts of loss data as a pre-requiste," says John Kiddy, chief executive officer at Chase Cooper. "And secondly, the business benefits flowing from an AMA are initially unclear - except to get a lower capital charge. Whereas people can opt to use either the basic indicator or standardised approach to calculate the capital charge anyway. This has led to institutions getting the wrong view of AMA."

He continues: "Once people get over that barrier and actually start to use the AMA, all sorts of business benefits flow out of it, which are very clear to them. Some of the confusion maybe is because there has been so much written from an academic perspective. In addition, some of the software offerings have further complicated the issue. Many organisations who have not opted for AMA think that it is very difficult, very expensive and of no real business benefit other than trying to lower their capital charge.

"I believe some sectors of the industry have lost sight of the underlying reasons why AMA was created in the first place - it wasn't just for a lower capital charge," he adds. "It was to have a risk-based charge for capital with all the benefits that brings."

The results back up his observations. Among those respondents who are choosing not to do the AMA, 28.6% said they didn't think the operational risk framework would provide a business benefit, while another 26.2% said they thought they were unlikely to get a sufficient capital reduction. Some 11.9% said they believed the framework was "too complex" for their firm.

The survey also showed differences between current AMA firms and prospective ones in what they plan to incorporate into their capital calculation. Scenario analysis proved to be the strongest component of AMA calculations for those who currently have them, with 91.3% using them. Some 82.6% use internal loss data, while 65.2% use external data and less than half - just 47.8% - use control factors. In contrast, for firms planning to use the AMA, scenario analysis (76%) and internal data (72%) ranked lower. External data also ranked lower, at 62%, while control factors ranked a bit higher at 50%.

There were also differences in what types of data are helpful in controlling costs. Internal losses and scenario analysis proved the most useful for those who have implemented the AMA framework, while those who are planning to implement the framework believe that control factors, followed by internal losses, will be the most useful. Interestingly, those who had actually implemented the framework found control factors to be of no use whatsoever.

As well, there is variation in how firms are approaching the modelling itself. Those who have also implemented the AMA have consolidated around Pareto (52.6%), Lognormal (47.4%) and Weibel (21.1%) distributions for their impact distributions, and around the Poisson (91.7%) distribution for their likelihood models.

In contrast, those considering the AMA have not consolidated around a distribution, with Pareto at 27.3%, Lognormal at 43.2% and Weibel at 20.5%. Some 6.8% of firms are also considering the Gumbel distribution, which none of the firms who have implemented the AMA are using. On the likelihood side of things, only 74.2% are considering the Poisson, while 9.7% are looking at the Bi-nomial, and 6.5% are planning to use the Uniform distribution.

Says Kiddy: "That's quite interesting because at one stage, there were all sorts of academic papers around that suggested the extreme value distribution - another name for the Gumbel - was going to be the most appropriate distribution, but it clearly isn't."

"There is definitely a very strong feeling that the practice is very different to the planning," says Kiddy. "How AMA operates in practice, the benefits, and the factors that go into it are completely different to what people expect them to be. I think that's a marked trend for even the people who are not planning to use AMA, because their expectations of the benefits and the difficulties are significantly different when compared with the people who are using it."