Thursday, January 02, 2014

Confirmation bias - the NBN and facts

After my last post for 2013, in which I was taking as a given the need to build a fixed high speed broadband network, I've started 2014 engaged in a "twitter-sation" with a gentleman whose twitter description is "business broker".  I had completely forgotten that there is still a bunch of people out there who seriously believe that any investment in fixed broadband is a waste because it will be bypassed by wireless.

"The NBN is obsolete because of wireless" claim

The conversation started innocently enough with a question to me asking if I'd ever heard of apps and including this chart.

As I'd written about the idea of using technology to generate letters to MPs I was wondering if this was some cryptic hint that I should be looking for an app not a website.
The reply was that "Nobody is using cables. The world is going wireless at a rapid rate."
I replied citing the ABS stats on downloads per service (more on this later) This evoked the response "ABS stats? How about some global data that is up to date? You do know what Android is? " This was accompanied by the graph below.
Now I was finding all this interesting - in part because I was carrying on the twiiter-sation using my iPad and the home WiFi.  I asked my respondent what he was actually using and don't think I got a reply. 
Now the ABS Stats I forwarded were only to December 2012 because that was the set I found using my iPad and a quick Google search.  I realised that I'd actually put all the June 2013 data up in this blog post on 8 October. 
My interrogator seemed to place some great store in the use of international sources rather than domestic ones. I'm not sure I actually ever got the relevance of either of the two charts - especially since as they are both cumulative data sets the growth rates are relatively modest. Apps is a difficult one because there is no "limit" to it - but the Android devices showed a distinct slowing. 
But the question remains over what networks are the connections made.  The Cisco VNI global data and forecasts are as follows;
Global Fixed/Wired was 59% of total IP traffic in 2012, and will be 45% of total IP traffic in 2017.
Global Fixed/Wi-Fi was 49% of total Internet traffic in 2012, and will be 56% of total Internet traffic in 2017.
Global Mobile was 2% of total IP traffic in 2012, and will be 9% of total IP traffic in 2017.
When it is recognised that the period 2012 to 2017 sees the full explosion of mobile devices it is clear the per service download trends will continue for a while.
Attempts by me to use the comments of Telstra and Optus in support of the idea that fixed networks will still do the heavy lifting, and that the former CEO of Vodafone AU Bill Morrow was clearly backing the fixed line networks were met with derision.  The point that mobile networks rely on fixed was met with "here we go again… repeating same drivel. yes mobile tower uses fibre as backhaul. Why raise something so irrelevant?" I really didn't have the energy at that point to explain the mismatch between Cooper's Law and Moore's Law, that higher speeds like 4G don't magically make more spectrum appear so long term strategies are about smaller cells and traffic offload.
At the same time came the real stunner "As for Telstra - their 4G profits say it all!"  As I'm pretty sure Telstra doesn't even declare profitability separately between fixed and mobile, I am thoroughly confused about a report on the profits of 4G.  This, of course, did follow from the comment " look at financial & take up rate, then compare it to Telstra 4G."
So it is time to do some research.  What exactly is Telstra saying about 4G?
A quick review of the October 2013 Investor Day presentations reveals no separation out of 4G growth or profitability.  Pages 12 and 13 reveal that mobiles revenue in total is growing at 6%, but that margin remains at 38% - up a bit but not significantly different from 5 year trend of 36%.  Full year results announcement to June 2013 did include the comment "We now have sold 2.8 million 4G devices, including 1.9 million handsets and 200,000 tablets, 400,000 dongles and 300,000 mobile WiFi devices."  It noted that 27% of handheld devices were now 4G.  It did not record what percentage of time 4G devices used the 4G network or were actually roaming on 3G (at the time 4G only covered 66% of population.) It also noted that total CapEx of $3.8 billion "has enabled us to meet ongoing customer demand, support the accelerated rollout of mobile 4G LTE and meet ongoing delivery of NBN commitments."
So I don't really understand the assertion of high take-up rates and profitability.  NBN take-up is - despite incorrect reports to the contrary - actually very strong. (NBN Co's presentation to the JCNBN on 19 April 2013 demonstrates this.
My interrogator also questioned whether NBN Co should be happy with 2% completion after 5 years. Actual commencement of construction on the FTTP network only really commenced once the Definitive Agreements with Telstra went unconditional in March 2012.  The figures also pay no attention to the extent of build of the transit network (of which Telstra builds part), nor of fixed wireless and satellite. It is also amazing that the "2%" number endures despite there being an additional 100,000 premises passed from July to December.  And even the "Revised Outlook" from the Strategic Review only pushed the completion date to 2024 (Note also that if the FTTP build were to not cover the HFC areas the revised outlook would presumably see it complete by 2021).
Anyway the whole discussion of mobile as an alternative died when I got no response to my question on what device/network combination my interrogator was using. By the way for the record are screen shots for my iPad using WiFi and my 100 Mbps Telstra cable extreme and then using the Telstra 3G from home (to a Telstra testing point)

Simple observations: (a) given a choice for doing absolutely anything I will prefer the faster download speed - the difference is immense even on something as simple as refreshing the twitter app and (b) the upload is rubbish on both and its importance is often underestimated for ordinary browsing.
Corporate Plans and other reports
So the twitter-sation suddenly leapt into a third new territory with the comment "Why was hiding a report that values at NP Value of NEGATIVE $35b?" 
This sudden diversion conflated two issues.  The first was the assertion that the Government refused to release the Corporate Plan 2013-16 prior to the election.  This claim was made as recently as 18 November by Mr Turnbull in the Parliament, and on the same day Mr Albanese demonstrated why it was a false claim. (The draft was ultimately published on the AFR website after it had been leaked). As was revealed by NBN Co's CFP Mr Payne at a Senate Committee the Government had asked NBN Co to revise the draft that was submitted to reflect the effect of the suspenson of Telstra remediation work while new asbestos handling procedures were implemented.
The second was a report in the Australian on 16 November that was headed "Labor told of $31 billion NBN risk."  The existence of this report was not new - after all it was reported on by the AFR on 13 April 2013
The reports of the Lazard review contain all the essential information but the Australian version specifically has been crafted to mislead.
The review was commissioned by the Government to provide advice on the Telstra/NBN Co agreement (as reported by the AFR and in the actual story in the Australian), not on the overall project (as claimed in the Australian editorial of 18 November).  The key recommendation of the report was that the Government should seek lengthier non-compete clauses from Telstra. 
The Lazard report raised concerns about the effect of wireless substitution which was covered in the original Telstra agreement by a prohibition on Telstra promoting wireless as a substitute (as reported by the Australian). This was deleted at request of the ACCC (as reported by the Australian) but at the time all parties agreed it was actually unnecessary because to promote mobile as a "substitute" would be misleading and deceptive conduct.
The single biggest clanger in the original Australian report was to claim "Net present value calculations are done specifically to take the risks involved into account." They are not.  They are done to make it possible to compare two sets of future cash flows - in this case the "Telstra deal" and the "no Telstra deal" scenario.
The nub of the issue is included in one line in the Australian report "The advisers said the project had significantly underestimated the cost of its capital, and provided alternative figures, but in the end reasoned that this was a theoretical endeavour."
That's really it - to do the deal/no deal comparison Lazard simply used a much higher discount rate for doing the NPV calculation - in part because they wanted to use a lower discount rate to reflect the idea that the Telstra deal de-risked the project.  The $31 negative NPV was simply the outcome of doing that - and it certainly wasn't anything remotely approaching a '$31 billion risk."
Mr Turnbull himself fuelled the confusion over the Lazard report.  In the Parliament on 20 November Mr Turnbull noted that in December 2010 Minister Conroy had said the expected rate of return was 7 per cent.  He continued "The only problem was that this was one month—only one month—after the government received advice from Lazard, the investment bank, that the NBN project had a negative value of $31 billion. No wonder Australians lost faith in Labor. No wonder they are sick of their spin."
The implication is that somehow one could not say there was a rate of return of 7% if the Lazard report had found a negative NPV. This is not the case.
Discounted cash flow analysis is not something most people deal with - but business executives, economists, some accountants, and bankers would deal with it regularly.  The internal rate of return (IRR) of a project is the discount rate at which the NPV is exactly zero.  In a set of cashflows for an investment with upfront investment followed by eventual returns using any discount rate higher than the IRR will automatically generate a negative NPV. 
As the Australian article had noted this is exactly what Lazard had done.  No mystery. No new information. No new risk not quantified.
My interlocutor at various points in the twitter-sation had disparaged Australia's tech media as biased in favour of the NBN and had critiqued any pro-NBN comment as "busy publishing copy from Conroy's office." The latter was particularly hilarious because one high profile online journalist had observed to me that his apparent pro-Turnbull line had come about because Turnbull's office was far more useful in providing information for stories than Conroy's.
In particular he was praiseworthy of the Australian for reporting on the Lazard report when no one else had.  Technically he was wrong because the AFR had already done so in April. But the substance was that only the Australian had been provided with a leak in November.  And the article reads like they had the actual report, not just someone's comments on it.  But unlike the AFR with the draft Corporate Plan the Australian did not publish the report in full
It is mighty odd that the Australian has supported a call to overturn a long standing convention about cabinet submissions when it could apparently publish the information itself.  It is also quite odd because someone has clearly been hawking the story around since April about this report.
The protagonist as I now think of him was today asking "What? You still don't know about the report was sitting on since 2010? " This is my long form reply to that.
He quickly pointed to three new items in the Australian as "You guys are beginning to look like a doomsday cult! More articles on obsolete #nbn".  Note the source is again the Australian.
I resorted to this long form response because the twitter-sation was going nowhere. This really wasn't just because it is hard to discuss the kinds of issues here 140 characters at a time. Similar frustrating conversations can be observed in the comments sections on on-line publications - Delimiter has many examples.  And is just a mine of such conversations.
It reflects far more the principle of "confirmation bias."  The rather good article on Wikipedia defines this as;
Confirmation bias (also called confirmatory bias or myside bias) is the tendency of people to favor information that confirms their beliefs or hypotheses. People display this bias when they gather or remember information selectively, or when they interpret it in a biased way. The effect is stronger for emotionally charged issues and for deeply entrenched beliefs. People also tend to interpret ambiguous evidence as supporting their existing position.
 The original description was applied to a tendency of people to use tests of a hypothesis that confirm their hypothesis rather than ones that don't.  This is generalised to a biased search for information and a biased evaluation of information.
In the context of discussion of broadband policy the position of The Australian has been very clear. One interpretation of this though is that the paper adopts the position that it expects will sell more newspapers. Newspapers by definition are more attractive to people who are not "digital natives" and hence a broadband policy story that is likely to trigger a purchase is one that confirms the anti-technology bias. 
The reverse is largely true of mostly on-line sources.  That is it is a logical commercial response by each media to undertake its intermediation task in such a way to confirm the beliefs and hypotheses of the people they want to attract.
Indeed, there is a reasonable theory that the Murdoch media's political coverage in general over the last forty years has been more a matter of following public opinion than forging it. The correlation between the Murdoch policy stance and public opinion does not provide evidence of which way the causal flow operates. 
Confirmation bias is a real phenomenon; none of us is immune to it. However, having policy discussions in fragmented form such as Twitter or the grabs of the nightly news doesn't really make it possible to address policy issues in such a way that confirmation bias can be addressed.
In policy areas outside broadband the same facts apply.  People on the left will incant "social justice" and the like minded will find al the evidence why policy to help the less well off is justified.  But people of the right simply hear "culture of entitlement" or worse (see Marlene Goldsmith's Social Justice: Fraud or Fair Go).
The biggest problem with confirmation bias is that we don't know we are doing it.  The best way to avoid it is to take the very conscious effort to understand what the evidence being presented o the other side really says and write a response directly to it rather than to just quote an alternative statistic. 
I'm not sure I totally pulled that off in the above - but at least I think my response on the wireless question shows that I do not have a myopic view about the future of wireless devices.  Indeed there are deeper industry questions about whether only the three mobile operators themselves will be able to properly integrate the use of home based networks with mobile networks and thus create a strategic risk for the future of MVNOs like iiNet and TPG.
But the broadband policy debate is far more complex than protagonists on either side will admit.  In many ways it really is a little like climate change. Just as one cannot conclusively prove that the planet is warming due to the effects of human activity, because of the uncertainty of future applications one cannot absolutely assert what the full future benefits of ubiquitous fibre broadband will be. But just like with climate change one can make a risk-weighted decision on what the right choice should be.  In the case of climate it is to move to reduce carbon emissions.
The NBN Co strategic review actually found the cost savings of building to an all-fibre network in the future were relatively small compared to the overall cost of the project.  And this is with cash-flows assuming away most of the earlier high end use of an all fibre network.
The last time a major fixed network investment was made in Australia it was Telstra's decision to build HFC after Optus had.  It is well known that at the time Telstra expected the investment in HFC to be value-destroying - it had a negative NPV of about $4 billion.  But not acting had a bigger impact - negative $7 billion.  (see page 89 of Managing in Australia).  However, these financial forecasts DID NOT include revenue (or costs) for also using the HFC for Internet connectivity.  In April 2012 Telstra CEO David Thodey called Foxtel Telstra's best investment - which has to include the HFC investment.
And that ultimately is one of the core questions - what are the unknown upsides of building FTTP now. The one thing we do know is that if we don't build FTTP the upside will be zero. So what is the "prudent" decision?  Is it to forgo the upside - or make an investment and realise the upside.  Telstra's twenty year old HFC decision gives a clear guide. 
1. The Australian editorial referred to above opened with the following.
In April 2009, The Australian urged the Rudd government to produce the business case for what taxpayers could expect from their investment in fast broadband.
Despite repeated urging, Labor refused to subject the nation's largest infrastructure project to a Productivity Commission cost-benefit analysis. As did the Gillard government. The obvious reason they didn't want the numbers crunched for public scrutiny was the extravagance of their "Rolls-Royce" fibre-to-the-premises model.
This editorial reflects an ongoing confusion in Australia's media between a "business case" and a "cost-benefit analysis."  The business case was indeed published - it was the Corporate Plan 2011-14 published in December 2010.  That was subject to independent review by Greenhill Calliburn who found it to be of the standard you would expect from a major listed company. 
2. The ethics of leaks.
When a journalist is basing a report on a leaked document they have more leeway than with reporting on a public document to be selective in their reporting, or indeed to make factual errors that may stand uncorrected.
Neither the Journalists Code of Ethics or the Australian's Professional Conduct Policy makes any direct reference to the issue of leaked documents.  But they are featuring more highly in everyday reporting as digital technologies make it ever easier to copy and distribute whole documents. And they work far more effectively than "we have been told" stories.
Surely it is time that the policy be that the source document be published - not necessarily on the day of the story but at least with sufficient currency that the accuracy of reporting can be scrutinised.
3. The case of Israel
A former colleague just informed me of a December decision by the conservative Government in Israel to back "Digital Israel" which seeks to make all interaction between Government and citizens paperless.  That part is very much the same as "Digital First" which was announced by the Labr Government in June 2013.
Interestingly the article notes that "to ensure that everything works the way it is supposed to" the Government is also "subsidizing a fiber-optic network that will let Israelis surf the Internet at speeds of 1 gbps (1 gigabit, or 1000 megabits, per second) and more...By 2020, some 70% of the country should be covered with a total of 25,000 kilometers of fiber optics in place when the project is fully completed."
4. Apologies for the fact I accidently published this before it was finished.

No comments: