This morning, the IPPR held a press conference to release some early data on the upcoming elections from the latest Afrobarometer round. This is very exciting, because we haven’t really seen scientific, political polls in Namibia. In this post I’m going to briefly look at whether we can get something useful from these polls, or whether we should discard them as unrealistic.
A Primer on Polls
Many people mistrust polls. This makes sense because they intuitively seem untrustworthy: how can I call up a few hundred people and make claims about the whole country based on that? But it really works, even if it seems magical. George Gallup, the godfather of political polls, had a good analogy to explain why:
Sampling public opinion is like sampling soup: One spoonful can reflect the taste of the whole pot if the soup is well-stirred.
If you pick a good sample that represents the population well, it’s fine to draw broad, national conclusions from it. It seems unlikely but it works remarkably well.
So we know that Polls Work. But an individual poll can sometimes be off. This is why sophisticated forecasts — like Nate Silver’s famous one in the U.S. — often average several surveys for better results.
We don’t have that luxury here in Namibia. We will have to make do with just the Afrobarometer. So is it any good?
How Good is the Afrobarometer?
The last time the Afrobarometer fell before an election was in round 4, in 2009. Granted, that was a whole year before the election, and a lot of things can change in a year; but let’s see how close they were anyway.
There were two questions that directly looked to find out people’s party allegiance. One question asked respondents if they felt close to a particular party, and another asked people who they would vote for in the presidential election. I compare these to the actual election results from 2009 — presidential results for the question on presidential candidates, and National Assembly results for the question which asked which party they were close to.
“If a presidential election were held tomorrow, which party’s candidate would you vote for?”
|Refused to answer||10.42||N/A|
Do you feel close to any particular political party? Which party is that?
|Refused to answer||4.08||N/A|
It’s evident that the survey did remarkably well. Most election results are very close to the survey. There were a few discrepancies, but they can be easily accounted for:
1. The RDP did better in the election than in the survey, the CoD did worse. This makes sense. The survey was held in 2008, only a year after the CoD split at their extraordinary congress in Keetmanshoop, and only very shortly after the RDP was even founded. Many voters would not have heard of this new party, and as the RDP gained momentum it makes sense these voters moved over from the CoD.
2. SWAPOs support was undercounted in the survey by a lot. Some opposition-minded people will say this is evidence of election rigging; I don’t think so. The simpler explanation is that the survey gives people the option not to disclose their preference. If you add up the people who said they didn’t know, or didn’t want to share, you come very close to the differential. One tentative conclusion we could draw is that SWAPO is a default of sorts for those who don’t quite know how to vote, or who aren’t that politically interested. This is an indication that the opposition hasn’t done enough to make themselves look like an attractive option.
Should You Trust the Poll from this Morning?
It certainly looks like it’ll be a decent indicator of what’s to come. I haven’t seen the details — it might be that, because we’re so close to the elections and things are heating up, that more people refused to answer, or that opposition party supporters would be more reluctant to express their opinions. So the opposition vote could be undercounted. Then again, we saw SWAPO’s vote was heavily undercounted too. Nevertheless, the general picture shouldn’t be too far off.