Dear Editor,
In Stabroek News, October 31, 2011, ‘NACTA polls put PPP in the lead,’ PPP pollster, Mr Vishnu Bisram, claims that the results of his latest poll show that “the PPP is polling 48%, followed by APNU 25%, AFC 8%, JFAP 1%, other parties less than 1%, and 17% undecided or no response.” He also claims that “The poll randomly interviewed 500 voters to yield a demographically representative sample 45% Indians, 30% Africans, 16% Mixed, 8% Amerindians, and 1% other races.”
There is a bit of a non-sequitur here. While a random poll may, in theory, yield such an outcome, it is logically impossible in Guyana. Which community in the country contains such a unique combination of ethnic cohorts to “randomly … yield” the stated outcome? The ready answer, of course, is none. Suspicion increases when we note that the sample bears a curious resemblance to the 2002 census, especially for the major ethnic groups – East Indians, Africans, Mixed and Indigenous communities – suggesting that the sample is simply made to arbitrarily coincide with the census while serving as alibi and proof of its unquestionable source. (But more of this below.) While the poll may have “randomly interviewed 500 voters” it could not yield the “representative sample” suggested. For such an outcome to be possible the process had to be selective not random.
Mr Bisram has been under heavy fire for some time to reveal the procedures and methods of his sampling. This he has consistently refused to do. Normal, accepted practice suggests that for a truth claim to hold its own its procedures and methods must be shown to be repeatable and verifiable under varied circumstances and available for scrutiny by independent observation.
Thomas Kuhn (Structure of Scientific of Revolution) has been arguing since 1960 that the “normalization of a scientific practice” requires iteration and independent verification if its assumptions are to be accepted. Otherwise, it remains at the experimental stage, where its results are guesses at best. Without independent verification the claims are just that – claims. For them to be published as verifiable truth based on the insistence of the claimant is retailing ideological hubris.
In the meantime, since the publication of his poll, Mr Bisram has had occasion and opportunity to answer a number of his critics. He has, unfortunately refused to respond to a letter from Mr Rickford Burke (SN, November 3). Mr Burke asked three questions of Mr Bisram: a) the community from which his sample was obtained; b) the database from which he drew his sample and whether that database is reflective of the electorate; and c) the procedure he used in the conduct of the survey, ie by phone or in person.
The issue raised in the first question has been settled and while the third is of import we may set it aside for the moment to deal with the second issue: the data base from which Mr Bisram operates. His sample, as we note above, bears a close resemblance to the 2002 census which gives the major ethnic groups the following demographic density: East Indians 43.4 per cent, Africans 30.2 per cent, Mixed 16.7, and Indigenous 9.2 – accounting for 99.5 per cent of the population. That data base is now nearly 10 years old.
Using its figures for 2002 and the three primary conditions that affect population changes under normal conditions – mortality, fertility and migration – the Bureau of Statistics concluded, employing a High, Medium and Low Variant Assumption (varying degrees of increase when the three variables operate under normal conditions/expectation) that the projected population of Guyana in 2010 would be 787,517 (High), 784,894 (Medium) and 777,873 Low).
Moreover, the censuses of 1980, 1991 and 2002 indicate that the group that has changed most is the East Indian. East Indians declined as a share of the population from 394,417 or 51.93 per cent of the population to 326,277 or 43.4 per cent. Indications, based on fertility and migration, are that the same trends continue in which East Indians suffer a net loss of some 3,097 persons annually or .94 per cent. However, when we take the second of the two periods, 1991/2002 the loss is less dramatic and provides a more accurate estimate of the succeeding period (2002-2011). While 1980-1991 saw a decline of East Indians from 394,417 to 351,939, a total of 42,478 at a rate of 3,861 or .97 per cent annually, for 1991/2002 East Indians fell from 351,939 to 326,277, a total of 25,662 , 2332 or .66 per cent annually. If our figures are correct or nearly so it would suggest that East Indians have fallen by another six per cent to a little over 37 per cent of the national population. It would mean, then, that Mr Bisram’s sample for East Indians is off base by 8 per cent.
If we were to add the 5 per cent margin of error that he admits, we have an error that is now 13 per cent of the putative database.
In addition, if the East Indians are now only a little over 37 per cent of the population which other group(s) would have gained and by how much and how are they reflected in Mr Bisram’s sample?
But Mr Bisram’s problems do not end there. Gecom has been at pains to explain, to anyone willing to listen, that it considers the 2002 census to be contaminated.
An issue, we are told, that was pointedly made to the President. It is so, it claims, because the census takers were paid on the basis of the numbers of names submitted and not a process of rigorous verification.
The inducement was obvious – census takers in a rush to be paid substituted cash for authenticity. And if Gecom is correct where does that leave Mr Bisram and his polls?
And Gecom is not to be trifled with. It reopened the period of “claims and objections” upon the request of the ruling party and promptly added some 7,000 names to the voters’ list.
What Gecom has so far failed to explain is if its own process of verification was so rigorous, if not foolproof, clear of the contamination that infected the census, how did it miss the 7,000 potential voters?
Part of the reason may very well be that it does not have a database of its own. And since it has dissed the census it is forced to make-up one as it goes along – only too happy to indulge the ruling party when requested. The troubling question is why has Gecom failed to provide its own database if it finds the census contaminated? And a data base, we want to add, is not just a one-shot deal but a process that establishes trends and patterns that is a source of surety and confidence.
Postscript.
I was about to complete this piece when another letter from Mr Bisram (SN, November 16) announced another poll with the same predictable outcome. Now we are told the PPP/C leads with 50 per cent support of Mr Bisram’s sample. In the latter instance he seeks to address one question raised by Mr Burke: the method of interview was “face-to-face.” And then we are back to the old story – “560 voters representing the varied demographics of the population (44% Indians, 30% Africans, 16% Mixed, 9% Amerindians, and 1% other races).”
The only question here is what is it about these figures that Stabroek News finds so captivating to publish at such a sensitive moment and troubling circumstances?
Yours faithfully,
Rishee Thakur