Saturday, November 2, 2013

Why Daily Star's opinion survey is a waste of time and money

The National Public Perception Study (also called 'Opinion Survey') undertaken jointly by the Daily Star and the Asia Foundation, and published today in the paper's 12 page tabloid spread - and also providing the paper with its front page lead article - is, simply put, a waste of time and money.

This may seem harsh, but the survey is a classic example of a pseudo poll, dressed up as a survey that is representative of the country, but in fact is not and as a result has very limited value.

The poll should not be given anywhere near the kind of attention that should be given to the recent polls undertaken by Nielsen/Democracy International and by Prothom Alo - if indeed it should be given any attention at all.

This is why.

In order for a survey to have any value, and for the Daily Star/Asia Foundation to be able to argue that the results are representative of the whole country, the poll has to be done in scientific way - and most importantly to use random sampling in order to identify the respondents.

This survey does not do this.

The purpose of an opinion poll is to ask questions of a small sample of a population of people who can represent the attitudes, opinions, or projected behavior of all of the people from which the sample is obtained. In effect, the fundamental goal of an opinion poll is to come up with the same results that would have been obtained had every member of a population been interviewed.

The key to reaching this objective is a fundamental principle called 'equal probability of selection', which states that 'if every member of a population has an equal probability of being selected in a sample, then that sample will be representative of the population.'

So for an opinion poll to be undertaken successfully in Bangladesh, every adult Bangladeshi must have an equal chance of falling into the sample. How this is done is the key to the success or failure of any such opinion poll.

Whilst the Nielsen/DI (published in the New Age) and the Prothom Alo polls used methods to randomise the selection of respondents - the Daily Star/Asia Foundation poll did not use this method at all.

As a result, there is little reason to imagine that the results of their survey are in any way representative of the whole country - which is what the Daily Star projects as being the case.

The methodology written in the newspaper puts much effort into explaining how weighting was done to make sure that the respondents were representative of the urban/rural divide for example - but this makes no difference if the original sample is not randomly selected.

And can one even explain the rationale for dividing the 100 selected respondents in each district into 'ordinary citizens' and 'non-political elites'? Is this how the country is divided?

In fact the Daily Star/Asia Foundation has done a great disservice to those organisations/newspapers undertaking proper opinion polls in undertaking this survey.

Bangladeshis are only just beginning to get used to the idea of opinion polls - and whilst most people may not understand the science/statitics behind it, many have started to appreciate that properly done opinion poll can produce legitimate results

For the Daily Star/Asia Foundation to undertake a survey with such little scientific value confuses all of this - and diminishes the work done by the scientifically organised polls.

It is particularly inappropriate for them to have undertaken this survey at such a politically sensitive moment in the country's politics. As the results are pretty meaningless.

The odd thing about all all this is that the Daily Star has in the past commissioned proper polling organisations to undertake scientifically rigorous opinion polls, so one would imagine that the paper's management are aware of the reason why one does an opinion poll rather than the kind of survey undertaken by them this time. But obviously not.

I can imagine that the Daily Star might respond to this criticism by arguing that it never actually said that the survey was representative of the national population - saying that it only 'gathered impressions' from a cross section of groups within Bangladesh. However, if you read the front page article, that is not the impression that is given. The impression given to the reader is that this survey is representative of the country.

There are three other oddities about this survey

1. Why did the Daily Star/Asia Foundation poll ask the question, 'Which party will win' rather than 'Which party will you vote for'. These are two very different questions, and will undoubtedly result in different results. 

That 55% of people think that that BNP will win the election, and only 28% think that the AL will win says nothing about the actual voting intention of particular people - and it is unclear why one would ask that question.

2. Why did the newspaper do this survey in September when its sister paper Prothom Alo had commissioned its own opinion poll in the very same period?

3. Why did the paper not at least commission  a professional surveying organisation to do the work?


  1. Being a research professional for long, I can say that no speculative questions should be asked to a respondent. Prediction/speculation is the job the analyst, using the obtained data. "Which party will win" is that kind of a speculative question. And that speculative question is presented as the primary finding of the study, which disturbs me. Regarding sampling, its a big debate whether weighting should be done at the start or in the end. I won't go into that.

  2. The poll was conducted in September (no dates were mentioned) and published in November. Waiting for what?