Apples and oranges: Why you can’t compare AfterAccess survey data with supply side data

Posted by on October 11, 2018  /  0 Comments

Our AfterAccess data was recently cited in an article by the New Nation, following the release of the Bangladesh findings in Dhaka. The media report refutes the AfterAccess survey numbers that we presented, which specifically peg Internet penetration at 13% of Bangladeshis between the ages of 15 and 65. Meanwhile, the New Nation cites data from Internet World Stats which indicates that half of Bangladesh’s population are “Internet users.”

The Internet World Stats numbers and AfterAccess (nationally representative) survey data cannot be compared; to do so is simply inaccurate.

The former is based on ITU data (which is usually a summation of industry subscriber numbers which are submitted to the regulator, BTRC, who then submits them to the ITU). The problem is that this kind of supply side data only tells us how many active data connections there are in the system, not how many unique users (i.e., actual human beings) there are.

It is a fairly well known fact that multiple SIM use is common in developing countries. This is certainly the case in Bangladesh, which has the third highest multi-SIM ownership across all the countries surveyed as the AfterAccess survey data shows. Thirty-four percent of Bangladeshi mobile owners between the ages of 15-65 own more than one active SIM card. As the New Nation points out: The BTRC said that the actual number of mobile phone users would not be match to the number of populations as the telecom rules allow an individual to hold up to 15 SIM cards against his or her national ID card.

The AfterAccess survey solves this problem by surveying individual users (regardless of how many SIMs or Internet connections she has), thus capturing the number of human beings who use the Internet in a country, within the target age group (15-65 years). Since our method is rigorous – random sampling at every step of sample selection, with a sample size of 2,000 – AfterAccess ensures statistical representation of the 15-65 population with a 95% confidence interval and +/- 3.3% margin of error  in Bangladesh.

The difficulties in using supply-side data as a way of benchmarking industry progress have been stated many times over by us, including in the Dhaka presentation, and by others. The fact that neither the operators nor the regulator seemed to disagree or challenge the AfterAccess (demand-side) numbers when we presented the data to them earlier this month, underscores the legitimacy of demand side data such as this.

For all the same reasons one can’t equate the number of (individual) mobile subscribers with the number of active mobile SIMs reported in a country, one can’t equate the number of (Individual) Internet users with the supply side numbers reported by industry. Added to the problem of multiple SIMs, one connection might even be used by more than one person. Household connections and corporate connections, and the IoT aside, one would think this is a fairly obvious thought in a country with a much-celebrated history of shared mobile connections being a way of overcoming affordability challenges.

Just for completeness sake, the comparison of AfterAccess survey data with supply side data is problematic for another simple reason. The target age group that we focus on is the 15-65s. So when supply side numbers are divided on the total population base, the resulting ‘penetration’ numbers are also different from ours because the population base is larger. The fact that babies are counted in the supply side population base, but also that below 15s and over 65s are excluded in our target group is without a doubt less than ideal. Perhaps in an ideal world, demand side surveys would include the younger (below 15s) and elderly (above 65s) age groups, but in the real world research budgets have to be managed, and in the process trade-offs have to be made, to ensure what is captured, is an accurate reflection of reality.

(As an additional note, the New Nation also wrongly indicates that China was one of the AfterAccess countries surveyed in Asia. The countries surveyed in Asia are: Bangladesh, Cambodia, India, Myanmar, Nepal and Pakistan, and Sri Lanka is underway. The Asia report, which contains the available survey data from 18 countries can be downloaded here , and detailed methodology notes can be downloaded here.)

Comments are closed.