Estimating internet users: an evidence-based alternative in the absence of survey data

Posted on August 16, 2013  /  0 Comments

A paper based on work Roshanthi Lucas Gunaratne and I did over the past two years is finally published in a peer-reviewed journal, info. Here is an excerpt of the abstract:

Purpose – There are significant shortcomings in the current method of estimating the indicator “Proportion of internet users” by the International Telecommunication Union (ITU) in countries where demand-side data are unavailable. In the absence of demand-side surveys, governments calculate the proportion of internet users on the basis of the number of subscriptions and a multiplier, which leads to arbitrary values. Errors in such base indicators ripple through the system, causing significant errors in composite indicators, and should be minimised. The purpose of this paper is to propose a new evidence-based methodology, in the absence of demand-side surveys, to estimate the proportion of individuals using the internet.

Design/methodology/approach – This study found the drivers of internet use to be income and education. A regression analysis was conducted between internet users per 100 in countries that employed demand-side surveys and a new index based on the income and education components of the Human Development Index (HDI).

Findings – The regression analysis showed a strong correlation between the proportion of internet users in a country and the new income and education index. Using these data, a new methodology was developed that creates incentives for governments to conduct demand-side surveys and reduce the errors yielded by the previous methodology.

Social implications – It is proposed that this evidence-based estimation method be used by the ITU in the absence of demand-side surveys, instead of arbitrary multipliers applied to estimated subscription numbers provided by governments. If governments believe that their national circumstances justify higher numbers, they can conduct demand-side surveys.

The problem that we solved can be represented thus: Afghanistan and Burundi are quite similar in per-capita GDP, post conflict conditions, etc. Yet Afghanistan had used a multiplier of 500 to calculate the number of Internet users while Burundi used 13. If our method is used by the ITU, such perverse outcomes can be avoided.

Comments are closed.