Thursday, February 14, 2008

What If Pakistan Opinion Polls Are Wrong ?

A recent International Republican Institutes's public opinion poll in Pakistan has made big headlines. The results of this poll do not augur well for President Musharraf and his allies in PML-Q. While the science behind such polls is sound, there are limitations as shown recently in the US primaries. Similar polls conducted at the start of the primaries did not see the Obama phenomenon and wrote off John McCain. So the question is: What if the polls are wrong? When polls go wrong in the US, most Americans just recognize the limitations of the methodology and move on. But, would that be the case in Pakistan? Would this poll not give the ammunition to those politicians and parties who are already threatening violence if they do not win? I think these questions are worth pondering, especially if the consequences of the inaccurate poll result in significant loss of life in an already troubled nation.

Just to try and prepare the people for such as possibility, let me suggest that we approach the IRI and other polls like it with caution. Here are some of the reasons why:

1. The sample size is small and not truly random. The sample is biased by those participants willing to respond who generally happen to be more interested AND more available. I personally rarely participate in these polls because the pollsters usually call when I am about to start dinner at about 7:30PM. They usually do not call back.
2. People do not always tell the truth. They say one thing and do another. The most famous case of this kind happened in California gubernatorial elections in 1982, when many white voters said they would vote for Mayor Tom Bradley of Los Angeles who happened to be black but in fact voted for George Deukmejian, the white candidate for governor. This is now known as "Bradley Effect".
3. The national poll does not take into account regional, ethnic and class differences, the level of enthusiasm (likely voters) and multi-way races resulting in skewed seat allocation on constituency by constituency basis.
4. The way the questions are framed can change the outcome in significant ways. Asking someone "if you approve of XYZ?" is not the same as "Would you considering voting for XYZ? " As seen in this IRI survey, these two questions produced two different answers. While only 15% of respondents said they approve of Musharraf, 29% of respondents said they would vote for the coalition supporting Musharraf.

Let us all exercise caution in accepting these poll results as truly representative of the real vote scheduled for February 18. If these polls were representative of the general elections, there would be no need for holding large scale general elections in Pakistan or elsewhere.

No comments: