2 / 22
•
Clear Win
: A particular party/candidate is expected to win – unless there is a major social shift
that could not be detected at the time of study (50+% difference)
•
Advantage
: A particular party/candidate has an obvious advantage (33 to 49%difference) as
observed at the time of study across all the polling divisions
•
Mild Advantage
: A party/candidate has an observed advantage that can be erased (15‐32%
difference)
•
Unclear
: The sample is too small to note an advantage for any party/candidate in this
marginal/near‐marginal constituency from a single integrated snapshot – less than 15%
difference and/or have large proportions of undecided and/or ‘Not’ voters.
In order to achieve tremendous precision in each constituency,we would have needed accurate
proportionate sample for each geo‐cell (polling division). In integrated methodology this would mean
having at least 15 respondents in each polling division, resulting a massive sample. Only for 3 pull‐out
marginal seats were such efforts made. In cases where greater accuracy was needed,the numbers were
increased beyond 15. Each parish had a minimum quota of 60. This means that parishes with only 2
constituencies would automatically get 30 respondents per constituency. We had to have a minimum of
60 per parish as a unit of analysis because each categorical variable needed a minimum of 15. To
illustrate, in a parish we needed to be able to cut the sample in at least 2 variables (for example age and
gender; hence young {15} males {15} versus young {15} females {15}).Understandably, the sample size of
parishes was affected by the population size,given each constituency had a minimum of 15 respondents.
Note also that this study carries 2 types of votes – vote and might vote. When a respondent expresses
with certainty that he/she is going to vote only 2 responses can result: a vote (PNP or JLP) or ‘Cannot
say’. The latter is possible because the integrated instrument does not ask a direct question about the
voting intention of the respondent. When a respondent expresses uncertainly about voting he or she is
asked ‘Why’. This allows the researcher to place the uncertain respondent under ‘Might vote (JLP or
PNP)’ or ‘Not vote’ or ‘Cannot say’.
Might votes can shift. Consequently they are only worth a half of a
vote.Constituencies with high levels of ‘Might vote’ are indicators of problems on the ground. This is
the same for those with over 50% of ‘Not vote’ and/or ‘Cannot say’. These constituencies require
greater scrutiny than a single snapshot could afford. This will automatically throw them into the
‘Unclear’ category in the Prediction Table.
The sample of 1248 displayed in Table 1 suggests that there were slightly more males than females, and
more mature respondents than youth despite the attempts of the team to have equal proportions. The
sample reflects the reality of availability – males and mature persons are simply more accommodating
to discussions of a political nature in Jamaica. Note that there were twice more PN and UM respondents.
This was allowed to emerge from the constituencies. Data for the study were collected over a 7‐week
period ‐ between November 15, 2015 and January 10, 2016.
2