Political polling in Jamaica 2011

By Wilberne Persaud, Financial Gleaner Columnist

A fledgling researcher was tasked, as a practicum, with the job of determining aspects of health and social status of the younger cohort - age group 15 to 24 - of his community. Upon realising a pop/reggae concert was scheduled for the next Saturday, he felt fortunate. Surely, a big chunk of his group of interest would attend.

At intermission, he checked about 70 youths from the first few rows of the concert venue. They cooperated by answering his questions. They also agreed to visit the study group's doctor's office for routine health status checks.

Sixty attended the doctor's offices. Concert organisers confirmed 600 tickets sold at the gate. That's 10 per cent of his chosen population - not that bad he thought. What results did he get?

Of the 60, 42 were female and 18 male; 28 had vision problems, optimally needing correction for short-sightedness; 12 were mildly hearing impaired; 13 sported dreadlocks; 31 of the females had between one and three children; of the 18 males, 13 admitted to having children and seven claimed to be actively engaged in the children's lives and supporting them.

This is a wealth of information from these 60 respondents, but is this information representative of the target? The short answer is no.

Census data established the population in that age group in the community to be 12,000. His sample is therefore 0.5 per cent of the population of interest.

SAMPLE BIAS

That his sample is so small is problematic, plus the community is known to be fervently religious. A large proportion of the youth were in church at the time of the concert. No data from that group could be collected.

Worse, it is a dreadfully biased sampling of the already non-representative group originally chosen. His choice of respondents only from the front of the crowd introduces further bias.

Mildly vision-impaired youngsters would tend to be up front. They may not even know their vision was impaired. They don't visit the doctor for routine checks. Further checks might reveal they come from the poorer groups in the community. A range of correlations may exist as between income and social status, religious background, schools attended and so on. Much more care ought to go into selecting groups of people representative of the population of interest.

Here's another example of bias often overlooked. Steve is described by his neighbour as 'very shy and withdrawn, invariably helpful but with little interest in people or in the world of reality. A meek and tidy soul, he has a need for order and structure, and a passion for detail'.

His description was given to a group of US students, who were asked whether Steve was more likely to be a librarian or farmer. They chose librarian.

They overlooked known facts: there were more than 20 male farmers for each male librarian in the US. Because there were so many more farmers, it was almost a certainty that more 'meek and tidy' males would be found on tractors than at library information desks. Ignoring relevant statistical facts, they relied exclusively on resemblance as a simplifying heuristic - rule of thumb - to make their judgement.

In our thought apparatus, humans often rely on heuristics leading to predictable biases - systematic errors in prediction. This report comes from the work of Daniel Kahneman, Nobel laureate in economics whose research had significant impacts on fields as diverse as economics, politics and medicine.

CONTEXT AND TOO-CLOSE-TO-CALL

Polling potential electors for a picture of Jamaican election outcomes encounters all these problems and more. A pollster needs truthful responses to probing questions. Answers hopefully allow accurate conclusions about opinions held on political parties, leadership, programmes, past and expected performance and finally, how respondents would vote on the day of the opinion poll.

The late Carl Stone developed on matters political, an enviable and entirely deserved reputation for accuracy in reported polling results. Yet, Carl was neither statistician nor probability theorist. He would be first to admit.

Fact is, probability and statistical theory take us so far and no further in matters of human decision making. This, quite distinct from inanimate objects like a true coin which on the long haul, flips heads 50 per cent of the time, or tennis balls in our Lotto machines that behave consistently with true chance, or the probability of drawing King of Hearts from a true deck of 52 cards.

Our three recognised pollsters and a majority of analysts - ignoring obvious partisans - all claimed the election would be too close to call. Two of them, at the last moment found a trend towards the PNP. Anderson stuck out his neck in photo finish, while Bill Johnson was sanguine in his resistance to making 'predictions' based on his last sounding and the emerging trend. This didn't stop The Gleaner's political reporters and analysts from using his results to predict a marginal JLP victory.

Herein lies the difference between Stone Polls and today. First, Stone used a sampling frame that fit Jamaica's context of class, income, occupational background and spatial distribution of potential voters.

Second, in his enumeration methods he recognised the extreme polarisation of Jamaica's politics. So, for instance, to interview a vendor at Coronation market the clear preference was for the respondent to be alone with the enumerator - neighbours overhearing the interview lead to untruth!

In his discourses he spoke of urban, main road, rural and deep rural communities and electors. Whenever I indicated his method of sampling did not admit of the +/- 3% margin of error, he invariably dismissed the idea.

Third, he used his interaction - in rum bar, at street corner, etc - and knowledge of Jamaican society and polity to augment his polling numbers with extraordinary perceptiveness.

I well recall one Thursday morning in the 1970s, after his normal gruelling jog, his coming by my yard on Campus, to share his 'discovery', that Jamaica was likely to vote as one constituency. We spoke animatedly for close to an hour on the issues, after which he left asabruptly as he had come - his signature dark blue Volkswagen growling as he drove to Ring Road!

I know my editor shall have indulged me for this long piece. More anon, but consider this: the "worst thing anybody could ever do is to underestimate the intelligence of the Jamaican people". This is the wisdom of none other than Karl Samuda on the TVJ-RJR Ian Boxill Polls aired 11/09/2011. How ironic. Ian Boxhill, speaking through Skype on TVJ, informs the public that the poll giving a late bounce to the PNP was conducted using a sample from the electoral list. His previous polls did not - June plum vs ortanique. Using the electoral list as sample frame in Jamaica introduces intractable difficulties.

G2K and the JLP drank their own Kool-Aid. Funding, Facebook and Android ringtones allow viral dissemination of material, yet substance remains paramount.

Campaign backfired

The anti-Portia ad campaign backfired hugely. Jamaican voters rejected Golding's narrative of stepping down to youth, as they did the notion of youth, by definition, being new and different particularly in view of Holness' toeing Party line, obfuscating issues in interviews. Seaga's and Golding's endorsement of Holness was ill-advised in light of Gang of Five and 73 Dudus-Manatt dead.

People evoke memories and penetrate tissue papering over JLP leadership tensions that endorsement of Andrew Holness attempted to downplay. That 'intelligence of the Jamaican people', which Karl Samuda emphatically recognised, defied JLP backers, strategists and financiers.

My own study and surveys of Jamaicans, particularly rural folk, convince me there lives a robust, passionate morality - almost subliminally generated. It is found in 'day-for-day' rural effort on a square-chain of perhaps marginal land and the numerous pithy sayings such as, 'If de tap a de stream dutty, a wha you expeck a de battam!'

My conjecture is the polls didn't seek opinion on these and a range of other issues, answers to which should have given a clearer picture of voters' perception of the parties, hence their intentions.

It is not the 'science of polling' but rather 'scientists' using polling that failed!

Wilberne Persaud is author of 'Jamaica Meltdown: Indigenous Financial Sector Crash 1996'. Email wilbe65@yahoo.com.