Wednesday, October 06, 2010

How to deal with the calls for a CBA

Seems that Malcolm Turnbull is going to make an NBN CBA his singular crusade. He returned to it again today in the Fairfax press.

I've written previously about CBA and some of its failings. A CBA certainly won't do what Malcolm thinks it will unless a CBA is conducted on every piece of Government expenditure.

My suggestion that Malcolm is again taking advice from Henry Ergas receives support from Malcolm's suggestion that the CBA should investigate the cost and benefit of upgrading the existing network.

And if the national broadband network is the answer, what was the question? Given millions of Australians already have access to high-speed broadband and the public debate has been how to ensure all Australians have that access, why has the government failed to investigate what the relative cost of upgrading our existing telecoms network would be as opposed to trashing it and building an entirely new one?

This brings us to the question of how exactly a CBA would be conducted. The Senate NBN Select Committee famously recommended that Ergas be commissioned to do the CBA. However, the original piece of work from Ergas wasn't really subject to much analysis. He and I did trade blows in the pages of Communications Day but that isn't online (the CD original report is though.

For the record I'll note two of the problems with the Ergas analysis. The first is that it assumes away the benefits of higher bandwidth by arguing that content will simply be compressed - despite that not being the experience in the real world. The second is that in considering the alternative of upgrading the existing network it ignores the eventual investment in an FTTH network. In fact the expert panel report found that an upgrade of the existing network was;

unlikely to provide an efficient upgrade path to fibre-to-the-premises (FTTP), because of the high costs of equipment associated with rolling out a FTTN network that would not be required for a FTTP network (i.e. FTTN is not a pre-requisite for the provision of FTTP)

However, the fact that one CBA is flawed doesn't invalidate all CBAs. Senator Conroy's argument has been that the benefits (and to a degree the costs) are so uncertain that a CBA would be meaningless. However, that isn't necessarily the case. Michael Gordon-Smith (formerly of the ABA) is now the Australian Director of Hubbard Decision Research, the company built off the book How to Measure Anything. This makes the case that anything can at least be estimated and the process of estimation can be used to identify the size and consequence of uncertainty.

That would certainly be worth doing and is not the same thing as the business case nor the implementation study. In fact providing more information could resolve other simple issues like the confusion over how expensive it will be. Writing on Business Spectator Rob Burgess has compared a BT FTTP roll out to the NBN and questioned why ours is so expensive. In the end he answers it by noting that we are going to 93% coverage not 80 to 90. What he ignores is that the McKinsey/KPMG Implementation Study resulted in the 93% figure because that was what was achievable within the envelope of $43B.

My suggestion is that in conjunction with the NBN work that DBCDE should undertake a CBA of the NBN. However, the purpose of the CBA should be about managing risjk and uncertainty and ensuring optimal timing, not deciding if to build an FTTP network. Similarly there should be no slow down in any construction as it the initial construction will be the only definitive way to reduce some of the uncertainty.

More importantly it would suit new paradigm politics.

1 comment:

Anonymous said...

I'm posting a comment I received by e-mail from Doug Hubbard (I've turned word verification off - dont't know what the problem was). DH

I wanted to email you my comment on your blog. The window for posting a comment asks for a verification using an image but the image of the word I need to type appears to be broken in the window. Anyway, here is the comment I was trying to post:

Thanks for mentioning my work. You have made a great point in response to the Senator who is skeptical of CBA's on the basis that benefits are so uncertain as to render the CBA "meaningless".

In one sense, the Senator is correct. A *traditional* CBA would be a fairly limited value (if not quite meaningless) because the deterministic values do not communicate anything about their uncertainty. But as you might be aware of, uncertainty itself is something we include in decision models.

In fact, we need to quantify that uncertainty (which we do routinely) in order to assess the extremely important issue of risk. I'm unfamiliar with this particular investment, but it is unlikely that it is so uniquely uncertain that it exceeds the uncertainty of battlefield logistics, movie projects, new drug markets, and the value of cleaner drinking water - all things I've modeled before.

It would also be worth pointing out to the Senator is that the question is not just the errors of CBA's but the errors of the implied alternative - apparently unaided intuition. Research has shown that in the case of extremely uncertain estimates, experts do benefit from "decomposing" the problem (MacGregor, Armstrong, International Journal of Forecasting, 1994). That is, breaking an extremely uncertain variable down to another set of variables that might be slightly easier to estimate then mathematically combining them. The classic example is how the physicist Enrico Fermi would get his students to estimate the number of piano tuners in Chicago. At first, they assumed they could not possibly know that but then Fermi would ask them to break it down by estimating the number of households in Chicago, the share of households with pianos, the average time between tunings per piano, and the number of pianos a tuner could tune in a day. Invariably, they would end up closer than they were at first.

So a traditional CBA can definitely be improved upon but it appears that even the simple act of decomposition offered by a CBA does reduce uncertainty.