Digital Divide worse than appears in latest report

WIP Report 2017 User Types
User Types, from WIP Report 2017

Opinion piece: Bill Dashfield looks at the latest World Internet Project report and teases out some underlying figures. The article expresses his personal views.


User types, WIP Report 2017, phone survey only (11.45% non internet user, 88.55% Internet user
User types, WIP Report 2017, phone survey only

The just released 2017 World Internet Project (WIP) report looks at internet usage and attitudes, and also users and non-users. The users/non-users section says 94% of New Zealanders surveyed use the internet, and only 3% of people under 65 don’t use the internet. It says those on line are very active, whilst half the non-users have little interest in going online.

However, when examined closely the non-representative sampling method is strongly biased towards internet users. The real number of non-user Kiwis is double the 6% reported (at least 11%);  non-internet users have been disproportionately missed by the survey; the non-users sample is biased towards older people and so should not be applied to the whole of New Zealand.

This is important: research is used when discussing the divide, prioritising support and measuring effects. We mustn’t read more into the figures than the methodology supports.

Individual use versus household internet connection

Household internet connection almost always means that some adults in household are internet users, but the reverse is not true. Individuals may use the internet at work or on their mobile device, without a household connection.

Use only at work, or in libraries, or on the small screen of a mobile, does not constitute full digital inclusion – some of the benefits and ease of working are lost.  So a household connection is still key.  (A gray area is household access via tethering to an individual’s mobile device; the household is connected but only when that individual is there. The WIP report looks at access methods, but more analysis is needed to tease the numbers out.)

Around 20% of households are still not connected

A representative measure of internet non-users is in the Census, which in 2013 found 20% of households were not connected; this may still be a little low.  An ISP survey1 said that in 2013 18% of residences were not connected to the internet, close to the Census figure. Contrast this with WIP 2013 which found 92% of individuals surveyed were internet users2.

That ISP survey is repeated annually and shows little change since: 79% households connected (2014), 80% (2015), 81% (2016), 79% (2017).  So the Digital Divide has probably not significantly narrowed in the last 5 years.  Our partners on-the-ground experience supports this – they see a large, real and urgent need for digital literacy and connection.

The full monty

In today’s world being digitally literate and online is widely recognised as being important for success, whether educationally, financially, in access to knowledge or in social inclusion. The 20/20 Trust agrees – its vision is “New Zealanders fully participating in the digital world”. The current government says closing the Digital Divide is a priority, to get everyone able to access the benefits of being online.

But how big is the problem? How many people are offline in New Zealand, where and why?  Research data is vital: we need to know how to best prioritise support and measure its effects. As a digital inclusion and digital literacy organisation, the 20/20 Trust recognises the importance of accurate figures, especially those relating to non-users. The Trust has always had an in-built research component, welcomes external researchers, monitors digital inclusion research and advocates building a robust digital inclusion index for New Zealand, as do many others.

The World Internet Project survey (WIP), produces a wealth of data.  Every 2 years, since 2000, research organisations in 39 countries survey a sample of their population about their usage and attitudes.  Some questions are standard, to allow comparison between countries and identify trends over time.

Earlier this month, the WIP report for the “Internet in New Zealand 2017” was released.  It says: “New Zealand has an extremely high level of connectivity     Consistent with previous WIPNZ surveys, a connectivity rate of well over 90 percent was found.    There was a slight increase in the number of those connected but the change was more evident in the 25 percent decrease of those not connected from 8 percent down to 6 percent.” Later it says “a slight increase in (internet) user rate to almost 94 percent (93.8%).”

So 93.8% of New Zealanders use the Internet? And only 6.2% don’t?

Well, no, unfortunately.  The true figure for non-users is much higher.  The problem is that researchers* have to make tradeoffs between getting large numbers of responses for statistical validity, getting a representative (random) sample, examining different aspects of internet use, cost, time and practicality. When interpreting the responses to questions, it’s vital to keep the sample method and the sample size in mind.  That’s why many research reports show the number of answers to each question – shown as (n=xx); this shows when the sample size is dangerously low.

(* I have great respect for the Auckland University of Technology researchers, who have done other, ground-breaking, research. The online survey component was partly added to increase geographic coverage, get a better breakdown of online activities etc. The numbers they report are accurate for their sample, but give the wrong impression if applied to all New Zealanders.  They have acknowledged this as an excellent point, offered to do further analysis, and invited me to be part of the task force designing the next survey.  I look forward to doing so.)

50% of users asked if they used the internet were found from an online survey database.
50% of people asked if they used the internet were found from an online survey database.

Sample bias in ‘internet user’

For this ‘internet user’ question, the main problem is the sample makeup; if we forget this we’re won’t see the wood for the trees. Fortunately the researchers detailed this on page 2: they got 2,012 responses in total: 50% (n=1,004) from phone interviews,  50% (n=1,008) from active members of an “online panel database” from an online sample provider. That is, the second 50% are all people who have signed up to do internet surveys; naturally they will all answer “Yes”‘ to “Are you a current user of the internet”.

So the real percentage of internet users in New Zealand is the percentage answering ‘yes, within 3 months’ when surveyed by phone? (89%)  And 11% are non-users?

Well, no, unfortunately. Unpacking the phone sample (50% of the total) further,  60% were landline random dialing, 20% from random whitepage listings and the remaining 20% from random mobile dialling, 1/4 of whom (2.5% of the total sample) are known to be under 40 years of age.

What do we know about NZ phone and internet use? (Figure 6 in the WIP report doesn’t help much as it shows percentage of responses, not percentage of respondents – which we need for a multiple tick question.)

From our Computers in Homes programme research, experience and external research we know the digitally excluded are:

  • less likely to have a landline than the average New Zealander3
  • unlikely to be in the white pages –  many move house frequently, and are concerned about personal privacy
  • extremely likely to have a mobile phone4

So digitally excluded are likely to be significantly under represented in landline dialed samples, but fairly accurately represented in the random mobile dialing, unfortunately only 10% of the total WIP sample.  Even then they are likely to under-report as they are likely to have less trust, time and motivation to take part in surveys than average. Future reports may to well to use this as the predominant method to secure respondents, perhaps with an inducement.

So the true percentage of New Zealanders who are non internet users is likely to be significantly above 11%

Other factors to bear in mind when interpreting the WIP report:

  • using landlines as a major sampling method
      • biases it toward older people – they are more likely to have a landline5, 7; this would also slightly reduce the percentage found to be internet connected as older people are somewhat less likely to be internet users (this difference is reducing)
      • biases it towards those internet connected – as many internet connections are bundled with a landline6, although access by wireless connection and mobile phone tethering is increasing
  • using an  “online panel database” from an online sample provider as a major sampling method will
    • probably strongly bias the ‘frequency of use’ upwards – these are people that have self-selected as being willing to spend more time on the internet to do surveys
    • make the geographic related analysis of user/non-user suspect; one reason quoted for adding this sample was to improve geographic coverage8

The recent Census  data (when available)  will also report the number of households with an internet connection.  Unfortunately the Census this year was ‘Digital First’, had greatly increased non-participation rates, and will tend to under-count non-internet users even more than before9 – this will be covered in a future article.

Reasons for non-use 

The WIP looked at reasons for non-use, and what would help them become users.  Although biased towards older people, the key barriers/enablers found are the same as 20/20 addresses in its Manifesto for Digital Inclusion:

  • Affordable access to the internet (computer and connection)
  • Motivation to use the internet (interest, time and benefits)
  • Core digital skills (training/support/knowing how to use)
  • Trust in online services (security/protecting identity)

The 20/20 Trust’s programmes address these barriers.

WIP usage and behaviour analysis valuable

Most of the report deals with usage and behaviour of internet users and this is valuable, with the above caveat about frequency of use. Reanalysis of the data to show those sampled by phone versus those sampled online separately and compare the responses and numbers of each would also be very useful.


Bill Dashfield has been working in IT all his life and for community ICT and digital inclusion since the mid 1990s. He has carried out surveys and research into ICT for 20/20, government, business and for a Master in Information Management (VUW*) and has been (at different times) sponsor,  volunteer, Trustee and contractor for the 20/20 Trust.

Whilst managing CommunityNet Aotearoa in the mid 2000s for the Department of Internal Affairs he was part of the working group on the whole of government New Zealand Digital Strategy and advised on community IT project proposals.

He is also passionate about web accessibility, working with AccEase, and organises NetSquared Wellington, a meetup group for community organisations using technology for social good.

Footnotes and References

  1. Quoted in the InternetNZ  ‘State of the Internet 2017″ paper, page 4.
  2. The 2015 WIP “combined database was weighted, taking into account the survey design, incorporating probabilities of selection for each cell in the sample design and to correct for departures from Statistics New Zealand estimated proportions on several important parameters: age (grouped); gender; and ethnicity.” However, it still used Online surveys as well as CATI, and so will suffer some of the same problems.
  3. About 50% of Computers in Homes participants already had a landline (20/20 internal figures), fewer than NZ average (87%) in 2013 Census; a  2015 NZ survey ( ) found that 65% of NZ households had a landline
  4. A December 2013 Australian survey found mobile use almost universal in the 18-54 age bracket, 92% in 54-65, and 74% in the 65+. (In NZ, there have been more active mobile connections than people since 2007).  Landline use was strongly correlated with age, from 53% in 25-34 up to 93% in 65+.
  6. WIP Report 2017 75-80% of internet connections in the home are broadband, split between landline and UFB
  7. Older people are more likely to have a landline;  A 2017 survey of cellphone users found 20% did not have a landline (NZ Herald 2/3/2107 One in five Kiwis don’t have a landline phone anymore, survey finds). “The younger generation was leading the technological revolution, the survey found, with about half of millennials saying they didn’t have a landline connection, compared to only one in 10 baby boomers.”
  8. The percentage connection shown for the West Coast of the South Island (100%), and the Far North(93%) are warning flags here. We know from our programmes these figures are too high.
  9. The 2018 ‘Digital First’ census has seen an 80% increase in non-respondents (from 5.5% pop. to around 10%) which I largely attribute to the shift to Digital First online surveys. Even in 2013, some 20/20 partners reported, a high percentage of people they work with were non-respondents to the Census; they will be even more invisible now.