More on the ABS Internet Activity Statistics

The old adage referred to in my earlier post on these statistics was the "lies, damned lies and statistics" line. The point being that popularity wasn't just based on subscriber numbers but also on downloads.

The other interesting thing to do is to compare the ABS stats to other data sources.

This release is the first time in a while that the stats have actually given a distribution of all the ISPs by size. The interesting thing is to compare this to the number of service providers that are members of the TIO that claim to be ISPs or TISPs (the latter meaning they provide telephone and Internet services - like Telstra).

The chart below shows that the ABS numbers are well below the total number of merely ISPs let alone those service providers who are either ISPs or TISPs.

The chart might suggest significant under-counting by the ABS. Another comparison we can do is between the ABS fixed broadband number and the total fixed broadband services on either Optus HFC network or Telstra's HFC or copper network. The latter appear in Telstra's results as part of their retail SIOs with HFC, as the fixed BB wholesale, the ULL and the SSS numbers.

The closeness between the two data series is probably not really a good thing as the Telstra+Optus data leaves out Transact and others. But still - the gap is not terribly massive.

Understanding DSL is a little bit harder as you need to estimate how much of Telstra's Fixed BB SIOs are DSL and how many cable. I've used the crude approach of subtracting the Optus HFC number from the ABS total cable and fibre number to get a Telstra retail estimate. The Optus data is their on-net and off-net DSL base. The Telstra and Optus numbers are stacked, the ABS line is their total DSL.

A similar analysis can be done of broadband mobile wireless. Once again the Optus number is stacked on the Telstra one, while the ABS line is total market. The remainder is essentially Vodafone, resellers (except Virgin which is consolidated in the Optus number I believe) and the vividwireless base.

Finally the narrowband data can be compared, as it is below using the same approach.

Adding other providers is problematic as some don't release a lot of data, but I will expand the data set. I might not republish till after the ABS releases the December 2011 data in about April next year.

Novae Meridianae Demetae Dexter delenda est

New statistics - old adage

As I foreshadowed the ABS has today released the Internet Activity in Australia data that updates us a year on the ACCC report released last week.

Not that you'd know immediately from the media release which was titled "Mobile wireless connections more popular than DSL". This raises the important question of what we mean by "more popular". What they go on to say is there are now more wireless broadband services than DSL.

The graph they provide on their highlights page is a version of the proportions of internet connections by technology. Mine appears below.

As a proportion DSL looks like it is a technology the use of which is declining. The better picture is obtained by a graph of the actual service numbers by technology, shown below.

Indeed there are more wireless services than DSL, but not actually more than fixed when the 912 thousand cable and fibre services are considered. These are included in Other in the graph - which includes the 106 thousand satellite services. Interestingly both satellite and cable numbers declined over the last half year while DSL numbers continued to increase.

But popularity implies use not just existence. And it is here that the information noted with such relish in the ACCC's report becomes important.

The amount of data downloaded per service (the data is all data for the preceding three months) continues to rise rapidly for fixed broadband services but is relatively static for wireless. To put it into overall context - 254,947 terabytes were downloaded by fixed customers, while only 19,149 terabytes were downloaded by wireless.

I know which I would choose to define as popular.

Finally for this post I continue to wonder about the veracity of the ABS statistic. It is derived from data provided by ISPs, but the latest release has for the irst time in two and a half years revealed the number of ISPs by size category with a total of 421. The TIO counted 205 members as straight ISPs and a further 405 as Telephone and Internet Service Providers as at 1 January 2011.

I want to understand the source of that discrepancy.

Novae Meridianae Demetae Dexter delenda est

The re-reporting of statistics

The ACCC for some reason is not particularly quick at getting its reports on telecommunications out.

Last week it announced that its 2009-10 Telecommunications reports had been tabled in the Parliament.

Why the ACCC needs fifteen months from the end of the year to get this out hasn't been explained. They do have to do some work on price indices using data they can only obtain at 30 June, but it seems a very long time. Can you imagine running a business with data that is over a year old?

But the stunning part is the coverage. The ACCC's release said;

Subscription numbers for both fixed and mobile services suggest that while mobile take-up continues to grow, this is not necessarily at the cost of fixed line services. ... Fixed broadband remains the dominant technology for downloading data, accounting for 91 per cent of data volumes.

This download information was the bit reported by the SMH, ZDnet and iTWire.

However, the ACCC report is not new data - it is simply repeating and interpreting ABS data. Furthermore they are reporting only the June 2010 data while the ABS website released December data in April and advises the June 2011 data will be released in two days time - September 28.

For the record the per service download data and the total of fixed vs wireless broadband is shown in the chart below.

This demonstrates the core point that while wireless is popular it is mostly as an auxilliary not main service.

I look forward to updating the graph on Wednesday!!!!!


Novae Meridianae Demetae Dexter delenda est

Consumer behaviour, economics and research methodology

I've had some fun with the iSelect media release that purported that the "iSelect Broadband Report" had been released but wasn't.

I now want to talk about its central thesis, the comment attributed to Charlie Brown that;

Australians need to do their homework on broadband before signing up to a new contract. It doesn't need to be time consuming or difficult - if you use an online comparison service, you can evaluate plans based on your actual needs to find the best one for you.

This, of course, means use a site like iSelect. The size of the prize seems to be quite impressive - the whole $141,313,506.15 that they absurdly tell us will be saved. As I've elsewhere noted we are given no information on how this amount is derived.

We aren't even told over what time period this saving is made, but it looks to be an annual amount. Still a big number, but we are also told there are 7.5 million households subscribing to the internet. That's a saving of under $20 per household.

Economists who talk about anything other than theoretical non-existent markets will tell you that "search costs" are a significant factor in consumer behaviour, and that consumers will use other things, like brand and referral as substitutes. The iSelect research tells us that, on average, we shouldn't exert more than $20 worth of extra search costs in choosing broadband provider.

Their pitch of course is that their site significantly improves your search capability. Having had a look at it it is a proposition that certainly wouldn't be true for me, though it may be for other consumers.

More significantly the release reveals that "Not surprisingly, most people (66 per cent) say that reliability and maintaining connection are the most important features of a home internet service." This is something that price comparison sites provide no assistance with.

The release also says "Less than a quarter (24 per cent) understand how the speed of their broadband connection is measured (megabits per second), yet 86 per cent say speed is very or extremely important to them." This means the consumers are actually savvy, because the access speed tells you nothing about the contention ratios in the ISPs network. Users judge speed by how fast things happen on their screen not how many bits can flow from their modem to a network access point per second.

There also seems to be an unexplained discrepancy. The release cites the ABS statistic of 7.5M households with a non-dial up connection. Technically speaking this is reported by the ABS as 7.595 and should be rounded up and represents the bulk of the 9.739 million non-dial-up connections.

But the ABS also reports download stats. They report 191,655 Terabytes of data downloaded for the entire three months prior to the report date (31 Dec 2010). That is a grand total of 19 GB of download for the entire quarter for every broadband connection, or just over 6 GB per month.

Yet the iSelect release asserts "Australians are on average downloading more than 48 gigabytes of data per household per month." This is a massive discrepancy between the result from the iSelect research (presumably based on user self-reporting) and the ABS statistic (which is based on data provided by ISPs). Two potential causes of that discrepancy could be errors in the self-reporting of download usage or a very statistically non-representative sample.

The position marginally improves though if the vast gap between fixed and wireless downloads is included (see graph) but 31 GNB in a quarter is still less than 25% of the rate claimed by iSelect.

I'm not sure I have the energy for another round of questions to iSelect.

Novae Meridianae Demetae Dexter delenda est