Afrobarometer, a think tank whose recent survey showed that 64 percent of the Zimbabweans who participated in its survey trusted President Robert Mugabe and that 38 percent would vote for the Zimbabwe African National Union-Patriotic Front against 16 percent who would vote for the Movement for Democratic Change-Tsvangirai if elections were held tomorrow, has been forced to explain its findings.
Below is the explanation by two of its key people, Brian Howard and Carolyn Logan, issued yesterday.
Recent Afrobarometer survey findings in Zimbabwe created a firestorm of reaction in the news media and on social media – a fact for which we think Zimbabweans should congratulate themselves.
A willingness to confront and debate public-opinion data is a sign of democratic health.
A good proportion of reactions has been dubious and even angry at Afrobarometer for reporting certain numbers – in particular the finding that 64% of Zimbabwean survey respondents say they trust President Robert Mugabe “somewhat” or “a lot.”
Afrobarometer has a highly competent national partner in Zimbabwe, the Mass Public Opinion Institute (MPOI), which carried out the survey, released initial results, and is fully capable of explaining how these data were collected and analyzed.
As the pan-African, non-partisan network of which MPOI is a part, Afrobarometer does not involve itself in partisan debates, but we do value feedback, and maybe we can offer a few clarifying points for consideration as the public discussion continues:
Where does Afrobarometer get its data?
Afrobarometer is perhaps best known for its rigorous methodological standards.
We do face-to-face interviews with randomly selected respondents who represent the entire country.
Every adult Zimbabwean had an equal chance of being selected for an interview.
For those who question whether 1 200 respondents can represent the opinions of an entire country, or who suggest that we only interview supporters of a certain party, we encourage you to read about our methodology.
Sample sizes of 1 200 or even less are common in public opinion research all over the world.
And social scientists, statisticians, and evaluators of our work have described Afrobarometer’s methods as “the gold standard” for survey research in Africa.
Starting from the national census, we randomly select a list of census enumeration areas (EAs).
Within each EA, we randomly select sampling start points, and from there, we randomly select households.
Within each household, we randomly select an individual respondent (alternating between men and women to ensure gender balance).
Every region is represented in proportion to its share of the national population, as are urban vs. rural areas.
How reliable are Afrobarometer results?
Based on our scientific methods, including random selection of respondents, a sample size of 1 200 respondents can give a very good – but not perfect – indication of what all Zimbabweans think.
All survey research comes with a “margin of sampling error”.
With a population the size of Zimbabwe’s, the margin of sampling error for a sample with 1 200 randomly selected respondents is plus or minus 3% at a “95% confidence level”.
This means that we can be 95% confident that the results from our sample are within +/-3 percentage points of the results we would get if we interviewed every adult Zimbabwean.
So on the question about trust in the president, for example, there’s a 95% chance that if we interviewed every adult Zimbabwean, we would find that between 61% and 67% of Zimbabweans say they trust the president “somewhat” or “a lot,” and just a 5% chance that the true number is outside this range.
That’s not perfect – but it’s pretty good (and it is widely accepted survey research methodology).
How confident are we that respondents are giving us their honest opinions, and not hiding how they really feel due to fear or other concerns?
Public opinion surveys only produce valid data if respondents can for the most part offer their honest opinions, so Afrobarometer will not conduct a survey in a country where we do not think this is possible.
There are several indicators we look for to be confident that Zimbabweans are relatively free to speak their minds, including:
1) Are there examples of criticism of government, as well as praise?
2) What do respondents themselves tell us about how free – or not – they are to speak their minds? and
3) If we’ve conducted more than one survey in a country – as we have in Zimbabwe – how well do our findings match up with other evidence, such as election outcomes, over time?
We also make every effort to encourage respondents to feel they can speak freely.
We try to match interviewers with respondents by language, we go to their homes (instead of just calling them on the phone), and we use their language where possible (Shona, Ndebele, and English in Zimbabwe).
It’s true that Zimbabweans do not express full confidence in their own ability to speak freely: Only 51% say they feel “somewhat” or “completely free” to say what they think, and only 15% feel “somewhat” or “completely free” to criticize Mugabe – these numbers are cause for concern and must be taken into account.
At the same time, 65% are willing to say that the government is doing “fairly badly” or “very badly” at managing the economy, and a remarkable 88% criticize the government’s performance in creating jobs.
In fact we find lots of evidence that people are willing and able to criticize their government during our interviews – including by saying that they are not free.
We’d also note that over the 17 years that Afrobarometer has been conducting surveys in Zimbabwe, our findings have generally tracked well with election results and other indicators.
Are we forecasting the 2018 elections?
No. We asked respondents in January-February 2017 how they would vote if an election were held the following day.
A plurality refused to answer (24%), said they wouldn’t vote (11%), or said they didn’t know (5%).
That’s more than said they would vote for the ZANU-PF (38%) or for opposition parties (22%).
Besides, a lot can happen between now and next year. So we are not offering predictions.
Why do almost two-thirds of Zimbabweans say they trust the president? What do they mean by “trust”?
Sorry, these are (great) questions we can’t answer yet.
We gather quantitative data on about 100 questions, on lots of different topics (democracy, economy, poverty, corruption, climate change, tolerance. …).
We analyze the data and make our findings and the data publicly available.
Some of these “What does it mean?” questions may be illuminated by more in-depth analysis of the current data set and/or by the addition of qualitative research done by others.
But a lot of this work of analysis, reflection, and discussion has to be done by Zimbabweans themselves – activists, policy makers, journalists, citizens.
Brian Howard is Afrobarometer’s publications manager and acting communications coordinator. Email: bhoward[at]afrobarometer.org.
Carolyn Logan is deputy director of Afrobarometer and associate professor in the Department of Political Science at Michigan State University. Email: clogan[at]msu.edu.