Reliable Brokers
Online Investing
Alerts & Analysis
Easy Trading

Polling and public opinion

Opinion polling for governance remains under-developed in Bangladesh

Update : 31 Dec 2018, 12:00 AM

I have recently been engaged in opinion polling related to the election.

The outcome of the polling was that the AL would receive between 57% and 63% of the votes cast in a free and fair election. The polling does not provide an answer to the question of the number of parliamentary seats that would be won.  

A separate model was used to estimate the outcome of seats based on the total votes cast and the past performance of the political parties in the 300 constituencies. This article discusses the polling, not the model linking the seats won with the popular vote.

The essential role of polling in a democratic system

In a modern democratic system, polling of the population and the voters is an essential element of good governance. The idea of the wise man making decisions based on his wisdom and knowledge of the good is a dangerous myth. 

Politicians use opinion polling to guide governance and also to figure out what they should stand for in an election.

A good MP represents his constituents so it is essential to understand what are their problems, what do they believe in, and what alternative solutions receive public approval.  

At this time political opinion polling is an underdeveloped business in Bangladesh. Politicians make little use of it as most believe in their superior knowledge, so there is no need to spend money to find out what they believe that they already know.   

This idea that Parliament should reflect the views of the people stands in contrast to the growing power of the administrative state, the modern version of the myth of the good man as ruler. 

But in our modern complex world opinion polling is the great instrument available to the parliament’s members to understand and learn the wishes of their constituents.   

The integrity of the polling process

One of the ideas floated after the press conference on December 26, 2018, where RDC released the results of its work was  the accusation that there was “a conflict of interest.”  

The meaning of this statement is that a senior person in the AL commissioned this survey and the results were faked presumably to create the impression that the AL was doing better than it really is.

Of course, the RDC team rejects such an accusation. Journalists and talk show participants, with no attempt to understand the issues, leaped to the conclusion that there was dishonesty here.  

Really, as an obvious target of these accusations, I am particularly aggrieved. First, more than 200 persons worked on the program of surveys that RDC carried out.  

These men and women worked day and night to collect and process data in an unbiased way. They were deeply committed to telling the truth about the people’s opinion and did not let their personal views influence the technical job of conducting the surveys.   

Second, every step of the way is documented. Third, people should be careful about accusing other people of dishonesty. It is impolite when not based on any serious thought.

Let us go further into this. Newspapers and television enterprises all over the world sponsor polls. The New York Times and the Washington Post both very liberal pro-Democratic party newspapers sponsor such polls.   

Fox News and the Wall Street Journal on the other side sponsor such polls.  Does anyone see any conflict of interest there? I think not. The sponsorship by Independent Television when one of its most senior persons has strong AL connections is no different.

Issues in polling in Bangladesh

There are three important issues in all polling in Bangladesh and all require proper management if the poll results are to approach reality.

Assume I set out to measure the literacy of women age 18+ who are resident in rural areas. First, I have to define literacy -- for example, able to read and understand a passage from the newspaper of a defined level of difficulty, and then define rural.  

I must find a way to draw a sample of this rural, female population so that every woman has an equal chance of being selected. This is not hard to do to end up with say 200 villages identified by random sampling all over the country and I intend to select five women in each village 18 years or more and interview them. 

This gives a sample size of 1000, enough for a reasonably accurate result. The problem is how to select the five women from a village after I have selected the villages. 

The way to do this, in theory, is to make a list of all of the women 18+ in the village and select randomly five from this list. I say in theory, as this is a very difficult and time-consuming job.  

One can reduce the area to choose the women making the task of preparing the list simpler. But that reduces the accuracy of the survey for the sample size.  

Most survey work in Bangladesh handles this last step in a sloppy way with claims of managing that are not actually carried out. It takes too much time and costs too much. 

The second part of the sampling problem is to weigh the results so that the actual sample drawn reproduces the population in as many characteristics as possible. Many polls in Bangladesh do not do a good job of adjusting the results with proper population weights.  

The second problem is finding out how to administer the literacy test in a manner that gives the woman responding to the enumerator confidence, to tell the truth, and make her best effort for the test.  

The whole atmosphere around the interview must be congenial and the respondent must not fear the process.  If she does not feel comfortable she may not take the test seriously.

Finally, the enumerator has to do his work. If assigned 10 interviews then he has to go to each location he has been given and administer the literacy test.  He must not go to his living quarters and fill in a lot of questionnaires imaging answers or test results. This is all too common. 

How did we handle these three critical problems: The final step of the sample design was completed using the voter lists. The enumerator now has a specific name and address to locate to find the person to administer the questionnaire and complete the ballot.

The problem in political polling is uncertainty about the extent to which a person is lying. The respondent may be worried that, when she says she is going to vote for the opposition, her information will get into the hands of the government party and they will harass her.  

So, she thinks it more prudent to lie. The basic argument that sceptics make about poll results is that people are too frightened to tell the truth, so there is a bias towards the government party. This is a very difficult problem.  

Our approach is to create a ballot that allows the respondent to give her choice privately with no one observing and place it in a transparent ballot box mixed with lots of other ballots. It is not fool-proof but it is a big improvement over just asking, “for whom are you going to vote?” 

In the city council election, we obtained good results, that is our predictions were reasonably close to the outcome. In particular, that the AL would lose in Sylhet.  

We are modest about this. It is our view that with the ballot method we are getting much less misrepresentation and more honest opinion. Getting people, to tell the truth, is a continuing problem in all opinion surveys, even ones less sensitive than an election poll. Humility is good advice for those of us working in this area.

The third question -- the integrity of the enumerators’ work -- is another difficult one. Our approach is lots of field supervision by trusted persons and cross-checking. By cross-checking, I mean the office staff telephones many respondents to verify that the interview took place, and to check one or two questions.

All enumerators know this checking is going on. Nevertheless, we sometimes find that the enumerator has cheated and not done the work. The questionnaires we have obtained from such a person are thrown away. The RDC’s efforts to maintain quality responses from enumerators pays off in confidence in the validity of the information from the questionnaire.  

More on respondent misrepresentation. The argument made frequently to me by Person A says: 

1. “Your data is flawed as people will not tell you the truth.

2. I [Person A] know, from my experience and wisdom, how Bangladeshis behave and think. I talk to a lot of people and I know better than the survey.”

Our response is as follows:

Bangladeshis, and perhaps most human beings, tell someone what they want to hear. Most parents believe their children tell them the truth, despite unending experience that says this is wrong. Children try to tell their parents what they think the parent wants to hear.  “Have you done your homework?” “Yes”;  “Why are you late getting home?” “Studying at the school library.”  

So when the wise experienced Bangladeshi political analyst talks to people about politics they may well be telling them what they think they want to hear, not the truth.

Political science has long been philosophy or smart observer insight. As a result, for hundreds of years political science was not science, but rather, well written wild guesses with no scientific basis.  

In the past 60 years, political science has tried to figure out what is going on by carefully asking. This is done in opinion surveys, or by discussing and interviewing. These methods have driven progress in making political science more scientific.

The cavalier rejection by otherwise intelligent Bangladeshi elites of a scientific approach to political science masks a very dangerous tendency -- namely to reject science as a way to know things. Ask around if people believe heavy objects fall faster than light objects.  

That is obviously true, common sense and my superior ability to think about such abstract issues guarantee that. But of course, it is not true. Galileo demonstrated that objects fall more or less in the same way. The Catholic Church put him under house arrest for believing such a silly thing. 

Views of opinion polling for voter choice need to be informed by real concern about improving methods. But we should not have to hear the educated opinion based on nothing but a pompous ego.  

The argument that we do not know is a valid one; the argument that a person ’s judgment and experience are superior to a scientific approach is I would call a “Trumpian” argument, although he has plenty of predecessors.

The survey talked to 2,200 persons; related surveys talked to 53,000 persons. This is young and old, country and city, rich and poor, male and female.  

Randomness meant that RDC reached the full complexity of Bangladeshi life experience. The introspection of the most brilliant mind cannot match the depth of coverage of proper probability samples.

To understand what goes on in Bangladesh as a means of improving policy, opinion polling is essential.  Knowledge is tentative and must be built brick by brick in a scientific manner.

In a lighter vein here is a test

1. How many Bangladeshi Muslims believe that it is okay to stone someone for committing adultery?

2. How many rural households that have electricity have ceiling fans?

3. How much of remittances come through the bank and how much come through cash delivery? (Hindu system)

4. How often in days per week does a household eat fish?

5. What is the most important channel for learning about politics? 

6. How many husbands physically beat their wives? 

7. How many people are unemployed? (they want a job but they do not have one)  

All of these have been answered, say +/- 3% by different surveys. If the answer is close to 50% the error is as stated. But if it is 10% or 90% then it is +/-1%.

Here are the answers and a few modest conclusions

1. 40% - don’t get caught

2. 99% - a great improvement in comfort and health

3. 50% - the official macroeconomic picture is out of whack.

4. 6.5 days - produce more fish for the poor households

5. Talking to friends - bad news for journalists

6. 70% - look at the guy sitting next to you

7. Less than 5% - everyone has a job, everyone wants a better one.

Of course, you knew all of that.

Understanding political developments is essential for the way forward to build a successful democratic society.  Not making judgments but understanding what goes on in the society is essential and the opinion poll is one of the most powerful tools for accomplishing this. 

Forrest Cookson is an American Economist.

Top Brokers