Dorab Sopariwala splits the essential challenges in politics-related research into the ‘intellectual’ and the ‘logistical’: The first intellectual challenge is converting votes into seats. This problem is faced in most countries that have a first-past-the-post system. A party that wins 15 per cent of the votes may land up with only five per cent of the seats in this system. In a country like India, with its large number of parties and candidates, there are candidates who win elections with fewer than 30 per cent of the votes cast. In addition, researchers often use the swing from the previous election to assess the seats likely to be won by the party in the forthcoming elections but this also is a problem because a party that was an ally last time is an enemy this time, thus making comparisons very difficult.
The second challenge is to do with the turnout. In an opinion poll, the voter is asked whether he intends to vote in the forthcoming election and those who do not intend to vote are not questioned further. In India, 90-95 per cent of the voters state that they are likely to vote – come floods, rain or civil unrest. However, on polling day, the turnout is just around 60 per cent. Now, the researcher has to hope and pray that the 30 per cent who stay at home are, in terms of voting choice, like the 60 per cent who come to vote. This is the most dangerous assumption the researcher is required to make.
The first among the logistical challenges is due to low tele-density; Indian pollsters do not have the advantage of conducting polls by telephone. They have to go to way out and fairly inaccessible villages to conduct their polls, which makes the job much more taxing than that of a pollster in a developed country. Moreover, this means that polls take a longer time (than in the West) to conduct.
Second, is the method of selecting the respondent. Having selected the constituency, the researcher can select a sample from the electoral rolls or based on the market research “random walk” principle or carry out street corner interviews. Each of these methods have implications in terms of methodological rigour, time taken to conduct the poll and the cost.
Third, in a national poll, over half the sample is illiterate or semi-literate. How do you get through to them? For voting preference questions, cards with symbols of parties are shown; these are not so much of a problem. However, they cannot read cards on which multiple options are written down. If you read out five options, will they remember the first by the time the fifth one is read out?
Times - India Times.com