Chart of the week #22: Trying to explain the low turnout

Last week, I drew attention to the extremely low turnout figures recorded at the Tanzanian 2010 presidential election. This week, I thought I would look at whether these turnout figures vary between different sections of society.

For this, I have turned again to the 2012 Afrobarometer survey, which asked respondents whether or not they voted in 2010, and if not, why not.

Overall, 81% said they voted. This is much higher than the actual turnout as reported by the National Electoral Commission, which was 43%. And the Afrobarometer methodology explains that the survey included respondents from the age of 15 upwards. Given that only those aged 20 and above in 2012 would have been eligible to vote in 2010, that means a considerable portion of the Afrobarometer sample were not eligible in 2010.

It’s not clear why the Afrobarometer figures are so massively different from the official NEC data. But I can suggest three possible reasons:

  1. People didn’t want to admit to the person asking the questions that they hadn’t bothered to vote.
  2. The Afrobarometer sample is somehow biased in favour of more politically-engaged people.
  3. There is a problem with the NEC data.

I think it’s very unlikely that either reason 2 or 3 is correct. Both Afrobarometer and NEC are very careful about these things. But reason 1 doesn’t sound convincing either – it would mean that half the people who answered the question lied?

Anyway, I thought I would carry on to see whether the numbers revealed anything interesting, though remembering that the data here looks unreliable, so we should be very cautious with interpretations.

First, I looked at whether reported turnout (according to Afrobarometer) varied between different groups in society. If people in rural areas reported having voted in much greater numbers than people in urban areas, for example, that would begin to give us clues about possible reasons for the low turnout.

I looked at gender, urban / rural, age, education level, and region:


Aside from a few exceptions, there’s not much difference to see here. The only really interesting exception is on age: that young people (aged 15-29 in 2012) reported voting in much lower numbers (61%) than older people. It looks to support my previous idea: that low turnout must have somehow have something to do with young people.

However, this data could be the result of people being too young to vote, so I looked at the breakdown of the reasons given for not voting:

Some (11%) in the 15-29 group did say that they had been too young to vote, and the same number said they had not been registered, which could again be because they were too young. But there was also a bigger group of younger people who said they hadn’t voted for some other reason (18%) than was the case for older people (11% and 4%).

Nevertheless, it’s unconvincing. The difference isn’t very big, and the lack of certainty in the data means I’m not prepared to draw any conclusions in this case.

One thought on “Chart of the week #22: Trying to explain the low turnout

  1. Pingback: What have you missed since June? | mtega

Comments are closed.