"Table III shows 14 constituencies which the PNP won over the JLP in 1997 by margins ranging from 34 per cent to 96 per cent. It is improbable that the PNP could lose any of those seats based on the present poll findings." (DK Duncan, The Gleaner, August 2, 2002, B10) Polling -- the expedient of asking questions of a limited number of people belonging to a larger group of people in order to arrive at conclusions about the views of the larger group -- is either dicey or dumb. And calling the "limited number" a "representative sample" does not rescue the operation from being dicey or dumb though it may add a veneer of sophistication. Just think, carefully, and bear in mind that we are not dealing with non-personal atoms or molecules or other such entities but with the complex variables called people. Why do I say polling is a dicey or dumb affair? Not only because you are dealing with people but also because it is not clear how the views of a small group could sensibly be seen as representing the views of a larger group unless you already know what the larger group believes, which is the very thing the poll is trying to establish. What is the logically necessary or probable connection between the views of the smaller group and those of the larger group and how exactly do you establish this? The very notion of a "representative sample" is "question-asking" or suspect. In what way(s) could one defensibly say that the views of, say, 2,500 Jamaicans are representative of the views of a larger body of 1.6 million Jamaicans? What criteria can really establish "representative", if you don't know the views of the larger body? Maybe I just lack training in the social sciences but one book our daughter lent me to look at on this issue seems to admit a bit of what I am saying. The book is The Practice of Social Research, 9th edition by Earl Babbie. The book admits on page 184 that "the term representativeness has no precise, scientific meaning" but advises that "a sample is representative of the population from which it is selected if the aggregate characteristics of the sample closely approximate those same aggregate characteristics in the population". What highlights the dicey or the dumb in polling is what is said further: "...representativeness is limited to those characteristics that are relevant to the substantive interests of the study. However, you may not know in advance which characteristics are relevant.. Really now! Leaving aside the dicey or dumb nature of polling, we move to the reasoning employed by Dr DK Duncan in his use of some Stone polls. In his article, from which we culled the opening quotation, Dr Duncan was very careful in his use of language. There seems to have been a lapse, though, in the quality of his reasoning which is evident in the quotation and its context in the article. Table III, captioned "very safe seats for the PNP", shows 14 constituencies won by the PNP in the 1997 election and provides the percentage margin of victory of the PNP as well as the percentage of votes cast for the NDM. Assuming the figures are accurate, the central problem of reasoning in the quotation is the link Dr Duncan makes between the 1997 election results re constituencies and the 2002 Stone Poll results re parties on the basis, presumably, that the latest Stone Poll showed the JLP and the PNP "locked in a statistical dead heat". How do you logically link party standing in the polls with election results for constituencies when we do not have an equal number of voters in each constituency and when it is possible in Jamaica to win the popular vote and not command a majority of seats in the Lower House? The other issue of course relates to the nature of polls. They tend to be treated in Jamaica as having predictive or even prophetic potential so elections are called by parties in power when the polls are saying the "right thing". Some people decide to vote or not to vote based on what the polls are saying. Individuals and companies decide to fund campaigns based on what the polls are saying re a party's chance of winning. Almost all writers on the possible results of the forthcoming election have written off "third parties" because of the polls. What must never be forgotten is the cruel fact that polls have no predictive or prophetic potential. They are simply guessing devices in fancy clothes? Another little thing about polls, this issue of margin of error. If you assert a margin of error of approximately five per cent you are silently, it seems to me, also asserting a margin of accuracy in the region of 95 per cent. If polls are, or can be that accurate, we could save time and money by pruning the election polls, taking their results and just do away with the actual elections. Would any social scientist strongly suggest this? Philosopher Alasdair MacIntyre may be too caustic but he has a basic point when he says, "...The aim of the social sciences is to explain specifically social phenomena by supplying law-like generalisations... precisely the kind of law-like generalisations to which the managerial expert would have to appeal...if social science does not present its findings in the form of law-like generalisations, the grounds for employing social scientists as expert advisers to government or to private corporations become unclear and the very notion of managerial expertise is imperilled...as...it ought to be; for the record of social scientists as predictors is very bad indeed, insofar as the record can be pieced together." (After Virtue, 88-89). |
