Let us all be Bayesian

Let us all be Bayesian

On October 16th, 2017, the Washington, D.C.-based opinion polling research organization, Pew Research Center, released its latest research report that comes from a global poll on citizens’ views on democracy, among other things.

The report finds that Tanzania tops, alongside Sweden and India, in its citizens responding favorably to the question: “How satisfied are you with the way democracy is working in our country?” Specifically, 79 percent of Tanzanian respondents report being satisfied with the way democracy is working in our country while 18 percent are unsatisfied. Only Indians are less unsatisfied.

Further, supporters of the ruling parties, across all countries, are more likely than supporters of the opposition to be more satisfied with the way democracy is working in their countries. In our case, supporters of the ruling Chama Cha Mapinduzi or CCM are 18 percentage points more likely to say they are satisfied with the way democracy is working in our country than supporters of the opposition. The specific split is such that among CCM supporters, 86 percent say they are satisfied with the way democracy is working in our country (implying 14 percent of CCM supporters are unsatisfied), while 68 percent of opposition supporters are equally satisfied with the way democracy is working in our country (implying 32 percent of opposition supporters are unsatisfied).

Note that the above result is interesting for three reasons. One is that only about a third of opposition supporters are unsatisfied with the way democracy is working in our country. Two is that although there seems to be a partisan divide of 18 percentage points in our country, ours is not particularly large and is in the bottom half across all countries in the Pew sample. Admittedly, the countries whose citizens are relatively most satisfied with the way democracy is working in their countries, India and Sweden, respectively, with 12 percentage points and 16 percentage points, have smaller partisan divides. Ours, however, is smaller than Nigeria’s (38 percentage points), the United States’ (28 percentage points) and our northern neighbors, Kenya (20 percentage points).

When respondents were asked about the trust they have that their governments will do what is right for their countries, Tanzania, solely sits at the top of the heap with 89 percent of respondents reporting that they either trust their government a lot or somewhat trust them to do what is right for Tanzania. Of course, almost half of these 89 percent of respondents only “somewhat” trust our serikali to do what is right for our country. However, 48 percent of Tanzanian respondents trust their government “a lot” to do what is right for our nchi, and this is second only to Ghana where 51 percent of their respondents trust their own government a lot.

There is an economic explanation behind all of these views. For instance, respondents from countries whose economies are the fastest growing tend to report trusting their governments more than respondents from slower-growing economies. Interestingly for Tanzania, our respondents trust their government more than would be predicted from this relationship between economic growth and trust in government. And in fact, this over-prediction is highest among all countries in the sample (top right corner of the graph below).


Further, among those who say the economic situation in our country is good, 94 percent of them trust our government to do what is right for our country. This is second only to Indonesia (95 percent). Moreover, 82 percent of those who say the economic situation in our country is bad, still trust our serikali to do what is right for our country — the highest of any country, with the second highest being Ghana, where 64 percent of their respondents who say the economic situation in their country is bad and still trust their government to do what is right for their country.

Note that the divide between those who are upbeat and those who are pessimistic about the economic situation in our country is smallest of any country in the Pew sample (12 percentage points). This means that Tanzanians who are optimistic and pessimistic about the economic situation in our country largely agree that they trust their government to do what is right for our country.

What are the take aways from all of this?

There are the following five take aways:

  1. Tanzanians are satisfied with their democracy and most importantly, even opposition supporters are likely to report satisfaction with their democracy.
  2. The Tanzanians government is trusted by its people more than any other government.
  3. Fast economic growth and good economic conditions predict Tanzanians’ trust in their government.
  4. Tanzanians who are pessimistic about our economic situation still trust their government.
  5. Results on Tanzania of the Pew Research survey are largely internally consistent and highly correlated

Is all of this real or an artifact of bad research?

Chambi Chachage, the Harvard University PhD student in African Studies, penned a piece on his blog that was, on average, skeptic of these results. In it, he writes that the “wall of defensiveness from supporters of the regime” and the “dose of skepticism from Tanzanian critics” compelled him to “interpret Pew’s results in their own right.” He also questions what people mean when they say they “somewhat” trust their government and points to possible contradictions in respondents’ views across different sets of questions that are conceptually related.

In personal correspondence between Mr. Chachage and I on Twitter, he explained that what he does not doubt is the integrity of the Pew Research polling results but rather that we not use this one piece of evidence as a banner to vindicate this current administration’s policies. To extend his argument then would be to incorporate pieces of evidence in a sequential manner and only after each subsequent piece of evidence corroborates the other then and only then should we conclude from what the data is telling us.

Mr. Chachage, whether he knows it or not, is thus asking us to be Bayesian in our thinking. I agree.

Baye’s simple rule is as follows:

where E or “evidence” in our case are the Pew research results; H or “hypothesis” in our case is that “Tanzanians are unsatisfied with their democracy” and/or “Tanzanians do not trust their government”.

Now, Bayesian thinking forces us to evaluate the probability of whether Tanzanians are unsatisfied with their democracy given the Pew Research Results. To do so, we need to know our prior probability that Tanzanians are unsatisfied with their democracy. If one is to read both local and international news, this prior probability would be quite high, so that P(H) = 0.9. I am certainly guilty of adding to this noise that things are bad by penning pieces in the Washington Post here and here.

We further need to know how likely are the Pew Research results true that Tanzanians are satisfied with their democracy if Tanzanians are actually unsatisfied with their democracy. Then we scale all of that with the probability that the Pew Research results are themselves true and not some “fake news”.

Mr. Chachage, I think, will agree that this latter probability should not be considered equal to zero or that the Pew Research results are an artifact of the ruling CCM or current government bribing the researchers or that government may somehow have intimidated respondents or that respondents, for one reason or another, are uninformed about what “democracy” truly is and thus were responding favorably without adequate information — all or most of the 79 to 89 percent of them!

If one thinks the probability that the Pew Research results are false then they would make P(E) = zero in the above equation (or 0.00000001 because one cannot divide by zero). Those people need not read any further because dividing something by a smaller and smaller number will ultimately make the probability that Tanzanians are unsatisfied with their democracy given the Pew Research results or P(H|E) increasingly large so that it may become almost certain that Tanzanians are unsatisfied with their democracy given the Pew Research results. Hawa ni wakereketwa wa upinzani.

Likewise, those whose prior probabilities or P(H) that Tanzanians are unsatisfied with their democracy are close to 1 or complete certainty, are also wakereketwa wa upinzani because a high prior probability will also ensure that P(H|E) or the probability that Tanzanians are unsatisfied with their democracy given the Pew Research results approach certainty.

On the other hand, imagine a sensible outsider who reads the Washington Post, the Economist, and Bloomberg and has thus read the negative stories about Tanzania’s current administration. Then that sensible outsider, let us call her Maresa, will have a high prior probability that Tanzanians are unsatisfied with their democracy, so that P(H) = 0.9, for instance. Maresa is a sensible human being, however, and so has faith in the professional integrity in Pew Research Center research staff and thus will also place a high probability that their results are in fact true so that P(E) = 0.9, for instance.

Now, what value would Maresa, this sensible outsider, place for P(E|H) or the probability that the Pew Research Results are true that Tanzanians are, contrary to local and international elite popular opinion, satisfied with their democracy, if they are otherwise unsatisfied with their democracy as many elite Tanzanians think?

Before the reader thinks about this value, notice that Maresa up to this point is at a deadlock. In the above equation, the 0.9 in the numerator (amount at the top of the equation) will cancel out with the 0.9 in the denominator (amount at the bottom of the equation) so that the only entity that will determine what P(H|E) or the probability that Tanzanians are unsatisfied with their democracy given the Pew Research results is P(E|H) or the likelihood that the Pew Research results are in fact true so that Tanzanians are in fact satisfied with their democracy if our hypothesis that they are unsatisfied with their democracy is also true.

The contradiction that would emerge in Maresa’s mind would, holding constant the previous evaluations of all other entities in the equation, lead her to make P(E|H) = zero or very close to zero so that in the end P(H|E) or the probability that Tanzanians are unsatisfied with their democracy given the Pew Research results would be zero.

The conclusion of such Bayesian thinking would at minimum make us doubtful about stories exclaiming Tanzanians’ dissatisfaction and/or distrust with their democracy and/or government, respectively. “Supporters of the regime”, like myself, would go as far as saying that in fact, precisely because the evidence from the Pew Research results is counter to our expectations, we should update downward our prior probabilities that Tanzanians are unsatisfied so that we may say that it is unlikely that Tanzanians are in fact unsatisfied with their democracy.

Similar updating needs to happen for opponents of United States President Donald Trump. Given he won an election in which people’s prior probabilities of his winning was low, we must all update upward those prior probabilities for the next election because he may just in fact win re-election.

A penultimate point. We all have our biases. I have a pro-CCM and pro-Magufuli bias. This means that these Pew Research results jive with my biases, even if my prior probabilities were equally high that Tanzanians are unsatisfied with their democracy, as many opponents of the current administration believe. I am, however, transparent about my bias and I also hope to convince the reader that despite this bias, I have never failed to be critical of the ruling CCM or its administrations, previous or the current one.

Readers of VijanaFM will note that I have been critical of CCM and/or government in many previous occasions. I have equally been supportive of the opposition whenever the objective truth is on their side. My Washington Post pieces cited above are one set of evidence that I am not one to shy away from criticizing “my team.” Previous pieces on VijanaFM here, here, here, here, and here are also testament to my efforts to be as impartial as possible. This is not to say that I have not been partial at times and biased in my pieces, I have.

What I hope my rejoinder to Mr. Chachage’s piece does, if anything, is to first provide an opposing view to his and secondly to demonstrate that as Tanzanians, especially young Tanzanians, we can disagree amicably and argue as objectively as possible.

Lastly, and I could not resist saying so, Mr. Chachage’s description of us, “supporters of the regime” being “defensive” is an incorrect description of our (or perhaps just my) current predisposition. When one is ahead three-nil against an opponent in soccer, who may be said to be on the “defensive”? Exactly.

Constantine was born in Dar es Salaam and raised between Dar es Salaam, Nairobi and Lusaka. He enjoys history, comedy, and African live music.