Estimating the number of Internet users in Sri Lanka; Facebook accounts as data source

Posted on November 6, 2011 / 4 Comments

Helani Galpaya’s work and LIRNEasia’s research has been drawn upon for a newspaper column. The novel element we had never thought of is using Facebook as a data source:

One other metric is available to anyone, just go to facebook.com/ads and create an ad. It will tell you how many people your ad can reach. For people of all ages, that number is 1,126,020. That is, Facebook has 1.13 million users that claim they’re in Sri Lanka. Even if you lop off 130,000 as errors, it’s still over a million Sri Lankans on Facebook.

So What Now?
So, to come round the bend, my rough guess is we have at least two million people on the Internet, or about 10% of the population. The raw numbers say 1.3 million, so why round-up so much? Well, because connections are used by multiple people, and I think we must have more Internet users than Facebook accounts. A lot of people also use the net at work, and thus wouldn’t be counted.

4 Comments

jason
November 7, 2011, 1:04 am
How was the data gathered, what were the techniques used to calculate the figures?

I think its wrong to take the Facebook as a source to get the numbers of internet users in SL, many individuals may have several accounts.
1. Rohan Samarajiva
  
  November 7, 2011, 12:04 pm
  
  The answers to the first question are in the post.
  
  Indi.ca is leaving out 130,000 Facebook users out of 1.13 million. That seems a reasonable accommodation for the problem of multiple accounts. What would be better is a rationale for the correction. Why 130,000? Why not 200,000? Why not 100,000?
  
  The best solution would be a sample survey to estimate the extent of multiple accounts. This would be different in different countries. Once that is known, the correction factor can be applied.
indi
November 7, 2011, 2:34 pm
The data is all overlapping. FB users is more of a check than anything else. If the amount of connections is 1.6 and FB registers 1.1 users, then the total amount of users is probably more than the connections.

The final number is a guess, but the other numbers are from the TRC, GSMA or Facebook
Helani Galpaya
November 7, 2011, 3:02 pm
It seems ITU is thinking about this too. Susan Teltscher of the ITU has alerted us to their ‘Measuring the Information Society 2011’ report (availble at http://www.itu.int/ITU-D/ict/publications/idi/2011/index.html) which contains a box on the merits (mostly de-merits, it seems) of using FB data to estimate users (see Box 5.2 on page 126 of the report). According to the report, FB uses the IP address to determine the “country” of the user. Some people travel and they could be messing up the numbers. Some people have more than one account (some accounts are for businesses) so counting is tricky. Agreed with all of the above. However, short of having proper demand-side survey data (i.e. representative sample surveys, or a question in a country’s census), most other methods of estimating (including that of using FB data) is a second or third best compromise.