By Andre Haynes
The race composition of the sample used in the recent CADRES political opinion survey was not representative of the national demographic profile, but the pollster maintains that even if it had been the findings about electoral support would have been no different.
According to a report on the results of the survey seen by Stabroek News, 39.6% of the sample was African-Guyanese, 26.6% was Indian-Guyanese, 22.2% was mixed, 7.1% was Amerindian, 2.3% was Portuguese, 0.4% was Chinese, 0.1% was Anglo (White) and 1.8% was classified as “other race.” However, at the time the last census was conducted Indian-Guyanese made up 43.5% of the country’s population, African-Guyanese made up 30.2%, Mixed Race persons made up 16.7% and Amerindians 9.2%, while Anglo, Chinese, Portuguese and others made up slightly under 0.5%. “It would appear as though your newspaper has acquired some information that was not released to it by CADRES and I do not care to comment on any such speculative information,” CADRES’ Director of Research Peter Wickham informed Stabroek News, when contacted about the race composition of the sample.
CADRES had done polling for the main opposition PNCR prior to the last general election. It has not however released the findings of that survey, which was commissioned by one of the party’s donors. However, a source who is familiar with that survey told Stabroek News that it was “a disaster,” noting that it was based on a similar methodology that under-represented Indian-Guyanese in the sample. According to the source, the findings suggested a PNCR win, with the AFC and the PPP/C, following behind in that order. The source suggested that Wickham was someone “who did not understand Guyana,” and, in particular, “sampling in a multi-ethnic society.” The PNCR subsequently secured 33.4% of the vote, behind the PPP/C’s 53.39% and the AFC’s 8%.
The March 2010 survey found that the governing PPP/Civic continued to command the single largest bloc of voters but had lost major support since the last general election. It measured the PPP/Civic’s overall support at 38%, the PNCR at 31% and the AFC at 26%, representing a drop of 16%, 3% and a gain of 18%, respectively.
CADRES has stated that there was a +/-5% margin of error at the national level. However, it noted that the survey was not designed to render a similar level of confidence at the local level and therefore assertions made in respect of the regions could only be considered a guide to public opinion in those areas.
‘Design bias’
CADRES did not establish pre-set racial categories in which interviews were to be conducted, the report on the survey seen by this newspaper said. “The interviewers were instructed to focus on those areas where political support for both parties was most neutral and in some instances, these areas turned out to be racially ‘unrepresentative’ of the national demographic profile…” it said. It added, “This is more a design bias than a design flaw and the poll might therefore exaggerate the electoral swing away from the PPP/Civic somewhat (most Indian-Guyanese supported the [party]).”
According to the report, the “design bias” did not mean that the shifts in electoral support identified were incorrect; rather, it argued that the poll would demonstrate “definitive trends” which were perhaps at this stage “slightly muted.” Moreover, it emphasised that “if the poll were to have canvassed 17% more Indo-Guyanese, the result would be no different, since the PPP/Civic would not attract the support of all these persons.”
The survey attempted to capture the opinion of 1,000 persons at the national level and 1,074 usable questionnaires were used to prepare the report. The data was collected using face-to-face interviews in all 10 administrative regions and interviewers were assigned to areas which corresponded to polling divisions with a relatively “close” relationship between both major political parties. The interviewers were also assigned gender and age categories along with quotas, which were to correspond as closely as possible with the national profile.
The report explained that such an approach targeted the “most politically sensitive” areas in Guyana. Between 12 and 36 interviews were conducted in each area and the interviewers selected the respondents based on a one in three formula and conducted the interviews in or near homes with persons who were already registered or who planned to register before the next general election.
A breakdown of the sample showed that 48.5% of the respondents were male, while 51.5% were female, while 33% were between the ages of 18 and 30, 36.2% were between the ages of 31 and 50 and 30.9% were 51 and over. Additionally, 50.1% of the respondents were working, while other categories included house-wife/husband (17.2%), unemployed (15%), student (6.0%) and retired (7.8%). The breakdown by income group, meanwhile, showed that 64.7% of the respondents earned below $50,000 per month, 31.2% earned between $50,000 and $200,000 per month, 3.3% earned between $200,000 and $900,000 per month and 0.8% earned over $900,000 per month.
A breakdown of interviews conducted by region indicate that 33.5% of them were conducted in Region 4; 15.6% in Region 6; 13.2% in Region 3; 8.8% in Region 5; 6.6% in Region 2; 4.5% in Regions 1, 7 and 9; and 4.4% in regions 8 and 10.
The PPP/Civic has strongly criticised the survey, calling it misleading. CADRES, however, has stood by its findings, saying it was satisfied that they were “a fair and accurate assessment” of the fortunes of the various political parties
CADRES has said it has a reputation as the region’s “most reliable” polling organisation. In responding to the PPP/C’s criticism, it emphasised that it adhered to the strictest methodology, which was available in detail to those who might be interested. “This method has been vindicated repeatedly over the 20 years that we have been in existence,” it said, adding that the organisation continued to be led by an individual who lived in the Caribbean and was intimately familiar with its political complexity.
CADRES noted that it had been in the business of political opinion polling since 1990. Recently, it predicted the win by the Democratic Labour Party in Barbados in 2008, the victory of the incumbent Roosevelt Skerrit in Dominica last year, the incumbent UPP’s success in Antigua last year and the victory of the incumbent Denzil Douglas in this year’s polls in St Kitts and Nevis.
Additionally, Wickham confirmed that CADRES had been retained by the COP in Trinidad in an advisory capacity. COP was part of the successful People’s Partnership coalition. He said while polling was part of CADRES’ remit, it also provided other support to the initial negotiation process and thereafter to the campaign.